diff --git a/matlab/method_of_moments/method_of_moments.m b/matlab/method_of_moments/method_of_moments.m index 5d86fb93d..9d6736116 100644 --- a/matlab/method_of_moments/method_of_moments.m +++ b/matlab/method_of_moments/method_of_moments.m @@ -63,7 +63,7 @@ function [oo_, options_mom_, M_] = method_of_moments(bayestopt_, options_, oo_, % o set_all_parameters.m % o test_for_deep_parameters_calibration.m % ========================================================================= -% Copyright (C) 2020 Dynare Team +% Copyright (C) 2020-2021 Dynare Team % % This file is part of Dynare. % @@ -92,6 +92,7 @@ function [oo_, options_mom_, M_] = method_of_moments(bayestopt_, options_, oo_, % - [ ] SMM with extended path % - [ ] deal with measurement errors (once @wmutschl has implemented this in identification toolbox) % - [ ] improve check for duplicate moments by using the cellfun and unique functions +% - [ ] dirname option to save output to different directory not yet implemented % ------------------------------------------------------------------------- % Step 0: Check if required structures and options exist % ------------------------------------------------------------------------- @@ -133,9 +134,9 @@ if strcmp(options_mom_.mom.mom_method,'GMM') || strcmp(options_mom_.mom.mom_meth options_mom_.mom = set_default_option(options_mom_.mom,'bartlett_kernel_lag',20); % bandwith in optimal weighting matrix options_mom_.mom = set_default_option(options_mom_.mom,'penalized_estimator',false); % include deviation from prior mean as additional moment restriction and use prior precision as weight options_mom_.mom = set_default_option(options_mom_.mom,'verbose',false); % display and store intermediate estimation results - options_mom_.mom = set_default_option(options_mom_.mom,'weighting_matrix',{'DIAGONAL'; 'DIAGONAL'}); % weighting matrix in moments distance objective function at each iteration of estimation; cell of strings with + options_mom_.mom = set_default_option(options_mom_.mom,'weighting_matrix',{'DIAGONAL'; 'DIAGONAL'}); % weighting matrix in moments distance objective function at each iteration of estimation; % possible values are 'OPTIMAL', 'IDENTITY_MATRIX' ,'DIAGONAL' or a filename. Size of cell determines stages in iterated estimation. - options_mom_.mom = set_default_option(options_mom_.mom,'weighting_matrix_scaling_factor',1); % scaling of weighting matrix + options_mom_.mom = set_default_option(options_mom_.mom,'weighting_matrix_scaling_factor',1); % scaling of weighting matrix in objective function options_mom_.mom = set_default_option(options_mom_.mom,'se_tolx',1e-5); % step size for numerical computation of standard errors options_mom_ = set_default_option(options_mom_,'order',1); % order of Taylor approximation in perturbation options_mom_ = set_default_option(options_mom_,'pruning',false); % use pruned state space system at higher-order @@ -169,6 +170,7 @@ options_mom_.mom.compute_derivs = false;% flag to compute derivs in objective fu % General options that can be set by the user in the mod file, otherwise default values are provided +options_mom_ = set_default_option(options_mom_,'dirname',M_.dname); % specify directory in which to store estimation output [not yet working] options_mom_ = set_default_option(options_mom_,'graph_format','eps'); % specify the file format(s) for graphs saved to disk options_mom_ = set_default_option(options_mom_,'nodisplay',false); % do not display the graphs, but still save them to disk options_mom_ = set_default_option(options_mom_,'nograph',false); % do not create graphs (which implies that they are not saved to the disk nor displayed) @@ -200,7 +202,7 @@ options_mom_ = set_default_option(options_mom_,'optim_opt',[]); options_mom_ = set_default_option(options_mom_,'silent_optimizer',false); % run minimization of moments distance silently without displaying results or saving files in between % Check plot options that can be set by the user in the mod file, otherwise default values are provided options_mom_.mode_check.nolik = false; % we don't do likelihood (also this initializes mode_check substructure) -options_mom_.mode_check = set_default_option(options_mom_.mode_check,'status',false); % plot the target function for values around the computed minimum for each estimated parameter in turn. This is helpful to diagnose problems with the optimizer. +options_mom_.mode_check = set_default_option(options_mom_.mode_check,'status',false); % plot the target function for values around the computed minimum for each estimated parameter in turn. This is helpful to diagnose problems with the optimizer. options_mom_.mode_check = set_default_option(options_mom_.mode_check,'neighbourhood_size',.5); % width of the window around the computed minimum to be displayed on the diagnostic plots. This width is expressed in percentage deviation. The Inf value is allowed, and will trigger a plot over the entire domain options_mom_.mode_check = set_default_option(options_mom_.mode_check,'symmetric_plots',true); % ensure that the check plots are symmetric around the minimum. A value of 0 allows to have asymmetric plots, which can be useful if the minimum is close to a domain boundary, or in conjunction with neighbourhood_size = Inf when the domain is not the entire real line options_mom_.mode_check = set_default_option(options_mom_.mode_check,'number_of_points',20); % number of points around the minimum where the target function is evaluated (for each parameter) @@ -738,11 +740,6 @@ catch last_error% if check fails, provide info on using calibration if present rethrow(last_error); end -if options_mom_.mode_compute == 0 %We only report value of moments distance at initial value of the parameters - fprintf('No minimization of moments distance due to ''mode_compute=0''\n') - return -end - % ------------------------------------------------------------------------- % Step 7a: Method of moments estimation: print some info % ------------------------------------------------------------------------- @@ -760,27 +757,56 @@ end if options_mom_.mom.penalized_estimator fprintf('\n - penalized estimation using deviation from prior mean and weighted with prior precision'); end -if options_mom_.mode_compute == 1; fprintf('\n - optimizer (mode_compute=1): fmincon'); -elseif options_mom_.mode_compute == 2; fprintf('\n - optimizer (mode_compute=2): continuous simulated annealing'); -elseif options_mom_.mode_compute == 3; fprintf('\n - optimizer (mode_compute=3): fminunc'); -elseif options_mom_.mode_compute == 4; fprintf('\n - optimizer (mode_compute=4): csminwel'); -elseif options_mom_.mode_compute == 5; fprintf('\n - optimizer (mode_compute=5): newrat'); -elseif options_mom_.mode_compute == 6; fprintf('\n - optimizer (mode_compute=6): gmhmaxlik'); -elseif options_mom_.mode_compute == 7; fprintf('\n - optimizer (mode_compute=7): fminsearch'); -elseif options_mom_.mode_compute == 8; fprintf('\n - optimizer (mode_compute=8): Dynare Nelder-Mead simplex'); -elseif options_mom_.mode_compute == 9; fprintf('\n - optimizer (mode_compute=9): CMA-ES'); -elseif options_mom_.mode_compute == 10; fprintf('\n - optimizer (mode_compute=10): simpsa'); -elseif options_mom_.mode_compute == 11; fprintf('\n - optimizer (mode_compute=11): online_auxiliary_filter'); -elseif options_mom_.mode_compute == 12; fprintf('\n - optimizer (mode_compute=12): particleswarm'); -elseif options_mom_.mode_compute == 101; fprintf('\n - optimizer (mode_compute=101): SolveOpt'); -elseif options_mom_.mode_compute == 102; fprintf('\n - optimizer (mode_compute=102): simulannealbnd'); -elseif options_mom_.mode_compute == 13; fprintf('\n - optimizer (mode_compute=13): lsqnonlin'); -elseif ischar(minimizer_algorithm); fprintf(['\n - user-defined optimizer: ' minimizer_algorithm]); -else - error('method_of_moments: Unknown optimizer, please contact the developers ') -end -if options_mom_.silent_optimizer - fprintf(' (silent)'); +optimizer_vec=[options_mom_.mode_compute;num2cell(options_mom_.additional_optimizer_steps)]; % at each stage one can possibly use different optimizers sequentially +for i = 1:length(optimizer_vec) + if i == 1 + str = '- optimizer (mode_compute'; + else + str = ' (additional_optimizer_steps'; + end + switch optimizer_vec{i} + case 0 + fprintf('\n %s=0): no minimization',str); + case 1 + fprintf('\n %s=1): fmincon',str); + case 2 + fprintf('\n %s=2): continuous simulated annealing',str); + case 3 + fprintf('\n %s=3): fminunc',str); + case 4 + fprintf('\n %s=4): csminwel',str); + case 5 + fprintf('\n %s=5): newrat',str); + case 6 + fprintf('\n %s=6): gmhmaxlik',str); + case 7 + fprintf('\n %s=7): fminsearch',str); + case 8 + fprintf('\n %s=8): Dynare Nelder-Mead simplex',str); + case 9 + fprintf('\n %s=9): CMA-ES',str); + case 10 + fprintf('\n %s=10): simpsa',str); + case 11 + fprintf('\n %s=11): online_auxiliary_filter',str); + case 12 + fprintf('\n %s=12): particleswarm',str); + case 101 + fprintf('\n %s=101): SolveOpt',str); + case 102 + fprintf('\n %s=102): simulannealbnd',str); + case 13 + fprintf('\n %s=13): lsqnonlin',str); + otherwise + if ischar(optimizer_vec{i}) + fprintf('\n %s=%s): user-defined',str,optimizer_vec{i}); + else + error('method_of_moments: Unknown optimizer, please contact the developers ') + end + end + if options_mom_.silent_optimizer + fprintf(' (silent)'); + end end fprintf('\n - perturbation order: %d', options_mom_.order) if options_mom_.order > 1 && options_mom_.pruning @@ -802,8 +828,6 @@ if size(options_mom_.mom.weighting_matrix,1)>1 && ~(any(strcmpi('diagonal',optio fprintf('\nYou did not specify the use of an optimal or diagonal weighting matrix. There is no point in running an iterated method of moments.\n') end -optimizer_vec=[options_mom_.mode_compute,options_mom_.additional_optimizer_steps]; % at each stage one can possibly use different optimizers sequentially - for stage_iter=1:size(options_mom_.mom.weighting_matrix,1) fprintf('Estimation stage %u\n',stage_iter); Woptflag = false; @@ -849,15 +873,20 @@ for stage_iter=1:size(options_mom_.mom.weighting_matrix,1) end for optim_iter= 1:length(optimizer_vec) - if optimizer_vec(optim_iter)==13 - options_mom_.vector_output = true; + if optimizer_vec{optim_iter}==0 + xparam1=xparam0; %no minimization, evaluate objective at current values + fval = feval(objective_function, xparam1, Bounds, oo_, estim_params_, M_, options_mom_); else - options_mom_.vector_output = false; - end - [xparam1, fval, exitflag] = dynare_minimize_objective(objective_function, xparam0, optimizer_vec(optim_iter), options_mom_, [Bounds.lb Bounds.ub], bayestopt_laplace.name, bayestopt_laplace, [],... - Bounds, oo_, estim_params_, M_, options_mom_); - if options_mom_.vector_output - fval = fval'*fval; + if optimizer_vec{optim_iter}==13 + options_mom_.vector_output = true; + else + options_mom_.vector_output = false; + end + [xparam1, fval, exitflag] = dynare_minimize_objective(objective_function, xparam0, optimizer_vec{optim_iter}, options_mom_, [Bounds.lb Bounds.ub], bayestopt_laplace.name, bayestopt_laplace, [],... + Bounds, oo_, estim_params_, M_, options_mom_); + if options_mom_.vector_output + fval = fval'*fval; + end end fprintf('\nStage %d Iteration %d: value of minimized moment distance objective function: %12.10f.\n',stage_iter,optim_iter,fval) if options_mom_.mom.verbose diff --git a/matlab/method_of_moments/method_of_moments_check_plot.m b/matlab/method_of_moments/method_of_moments_check_plot.m index 24b3ebb61..ceb1a15b7 100644 --- a/matlab/method_of_moments/method_of_moments_check_plot.m +++ b/matlab/method_of_moments/method_of_moments_check_plot.m @@ -2,7 +2,7 @@ function method_of_moments_check_plot(fun,xparam,SE_vec,options_,M_,estim_params % Checks the estimated local minimum of the moment's distance objective -% Copyright (C) 2020 Dynare Team +% Copyright (C) 2020-2021 Dynare Team % % This file is part of Dynare. % diff --git a/matlab/method_of_moments/method_of_moments_data_moments.m b/matlab/method_of_moments/method_of_moments_data_moments.m index 15e756955..26bf7a4dc 100644 --- a/matlab/method_of_moments/method_of_moments_data_moments.m +++ b/matlab/method_of_moments/method_of_moments_data_moments.m @@ -16,7 +16,7 @@ function [dataMoments, m_data] = method_of_moments_data_moments(data, oo_, match % o method_of_moments.m % o method_of_moments_objective_function.m % ========================================================================= -% Copyright (C) 2020 Dynare Team +% Copyright (C) 2020-2021 Dynare Team % % This file is part of Dynare. % diff --git a/matlab/method_of_moments/method_of_moments_objective_function.m b/matlab/method_of_moments/method_of_moments_objective_function.m index b97c7725f..745f8570e 100644 --- a/matlab/method_of_moments/method_of_moments_objective_function.m +++ b/matlab/method_of_moments/method_of_moments_objective_function.m @@ -31,7 +31,7 @@ function [fval, info, exit_flag, junk1, junk2, oo_, M_, options_mom_] = method_o % o resol % o set_all_parameters % ========================================================================= -% Copyright (C) 2020 Dynare Team +% Copyright (C) 2020-2021 Dynare Team % % This file is part of Dynare. % @@ -232,7 +232,7 @@ elseif strcmp(options_mom_.mom.mom_method,'SMM') i_ME = setdiff([1:size(M_.H,1)],find(diag(M_.H) == 0)); % find ME with 0 variance chol_S = chol(M_.H(i_ME,i_ME)); %decompose rest shock_mat=zeros(size(options_mom_.mom.ME_shock_series)); %initialize - shock_mat(:,i_ME)=options_mom_.mom.ME_shock_series(:,i_exo_var)*chol_S; + shock_mat(:,i_ME)=options_mom_.mom.ME_shock_series(:,i_ME)*chol_S; y_sim = y_sim+shock_mat; end diff --git a/matlab/method_of_moments/method_of_moments_optimal_weighting_matrix.m b/matlab/method_of_moments/method_of_moments_optimal_weighting_matrix.m index 7dde93568..ad2db20f7 100644 --- a/matlab/method_of_moments/method_of_moments_optimal_weighting_matrix.m +++ b/matlab/method_of_moments/method_of_moments_optimal_weighting_matrix.m @@ -19,7 +19,7 @@ function W_opt = method_of_moments_optimal_weighting_matrix(m_data, moments, q_l % This function calls: % o CorrMatrix (embedded) % ========================================================================= -% Copyright (C) 2020 Dynare Team +% Copyright (C) 2020-2021 Dynare Team % % This file is part of Dynare. % diff --git a/matlab/method_of_moments/method_of_moments_standard_errors.m b/matlab/method_of_moments/method_of_moments_standard_errors.m index 7bfd321b5..459f583f7 100644 --- a/matlab/method_of_moments/method_of_moments_standard_errors.m +++ b/matlab/method_of_moments/method_of_moments_standard_errors.m @@ -29,7 +29,7 @@ function [SE_values, Asympt_Var] = method_of_moments_standard_errors(xparam, obj % o SMM_objective_function.m % o method_of_moments_optimal_weighting_matrix % ========================================================================= -% Copyright (C) 2020 Dynare Team +% Copyright (C) 2020-2021 Dynare Team % % This file is part of Dynare. % diff --git a/tests/.gitignore b/tests/.gitignore index 722a27d5d..275d8736e 100644 --- a/tests/.gitignore +++ b/tests/.gitignore @@ -50,8 +50,10 @@ wsOct !/ep/mean_preserving_spread.m !/ep/rbcii_steady_state.m !/estimation/fsdat_simul.m -!/estimation/method_of_moments/RBC_MoM_steady_helper.m -!/estimation/method_of_moments/RBC_Andreasen_Data_2.mat +!/estimation/method_of_moments/RBC/RBC_MoM_steady_helper.m +!/estimation/method_of_moments/RBC/RBC_Andreasen_Data_2.mat +!/estimation/method_of_moments/AFVRR/AFVRR_data.mat +!/estimation/method_of_moments/AFVRR/AFVRR_steady_helper.m !/expectations/expectation_ss_old_steadystate.m !/external_function/extFunDeriv.m !/external_function/extFunNoDerivs.m diff --git a/tests/Makefile.am b/tests/Makefile.am index fd050916a..cd60e0473 100644 --- a/tests/Makefile.am +++ b/tests/Makefile.am @@ -50,10 +50,14 @@ MODFILES = \ estimation/MH_recover/fs2000_recover_3.mod \ estimation/t_proposal/fs2000_student.mod \ estimation/tune_mh_jscale/fs2000.mod \ - estimation/method_of_moments/AnScho_MoM.mod \ - estimation/method_of_moments/RBC_MoM_Andreasen.mod \ - estimation/method_of_moments/RBC_MoM_SMM_ME.mod \ - estimation/method_of_moments/RBC_MoM_prefilter.mod \ + estimation/method_of_moments/AnScho/AnScho_MoM.mod \ + estimation/method_of_moments/RBC/RBC_MoM_Andreasen.mod \ + estimation/method_of_moments/RBC/RBC_MoM_SMM_ME.mod \ + estimation/method_of_moments/RBC/RBC_MoM_prefilter.mod \ + estimation/method_of_moments/RBC/RBC_MoM_optimizer.mod \ + estimation/method_of_moments/AFVRR/AFVRR_M0.mod \ + estimation/method_of_moments/AFVRR/AFVRR_MFB.mod \ + estimation/method_of_moments/AFVRR/AFVRR_MFB_RRA.mod \ moments/example1_var_decomp.mod \ moments/example1_bp_test.mod \ moments/test_AR1_spectral_density.mod \ @@ -835,6 +839,10 @@ particle: m/particle o/particle m/particle: $(patsubst %.mod, %.m.trs, $(PARTICLEFILES)) o/particle: $(patsubst %.mod, %.o.trs, $(PARTICLEFILES)) +method_of_moments: m/method_of_moments o/method_of_moments +m/method_of_moments: $(patsubst %.mod, %.m.trs, $(filter estimation/method_of_moments/%.mod, $(MODFILES))) +o/method_of_moments: $(patsubst %.mod, %.o.trs, $(filter estimation/method_of_moments/%.mod, $(MODFILES))) + # Matlab TRS Files M_TRS_FILES = $(patsubst %.mod, %.m.trs, $(MODFILES)) M_TRS_FILES += run_block_byte_tests_matlab.m.trs \ @@ -984,8 +992,10 @@ EXTRA_DIST = \ lmmcp/sw-common-header.inc \ lmmcp/sw-common-footer.inc \ estimation/tune_mh_jscale/fs2000.inc \ - estimation/method_of_moments/RBC_MoM_common.inc \ - estimation/method_of_moments/RBC_MoM_steady_helper.m \ + estimation/method_of_moments/RBC/RBC_MoM_common.inc \ + estimation/method_of_moments/RBC/RBC_MoM_steady_helper.m \ + estimation/method_of_moments/AFVRR/AFVRR_common.inc \ + estimation/method_of_moments/AFVRR/AFVRR_steady_helper.m \ histval_initval_file_unit_tests.m \ histval_initval_file/my_assert.m \ histval_initval_file/ramst_data.xls \ diff --git a/tests/estimation/method_of_moments/AFVRR/AFVRR_M0.mod b/tests/estimation/method_of_moments/AFVRR/AFVRR_M0.mod new file mode 100644 index 000000000..8e51ac513 --- /dev/null +++ b/tests/estimation/method_of_moments/AFVRR/AFVRR_M0.mod @@ -0,0 +1,299 @@ +% DSGE model based on replication files of +% Andreasen, Fernandez-Villaverde, Rubio-Ramirez (2018), The Pruned State-Space System for Non-Linear DSGE Models: Theory and Empirical Applications, Review of Economic Studies, 85, p. 1-49 +% Adapted for Dynare by Willi Mutschler (@wmutschl, willi@mutschler.eu), Jan 2021 +% ========================================================================= +% Copyright (C) 2021 Dynare Team +% +% This file is part of Dynare. +% +% Dynare is free software: you can redistribute it and/or modify +% it under the terms of the GNU General Public License as published by +% the Free Software Foundation, either version 3 of the License, or +% (at your option) any later version. +% +% Dynare is distributed in the hope that it will be useful, +% but WITHOUT ANY WARRANTY; without even the implied warranty of +% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +% GNU General Public License for more details. +% +% You should have received a copy of the GNU General Public License +% along with Dynare. If not, see . +% ========================================================================= + +% This is the benchmark model with no feedback M_0 +% Original code RunGMM_standardModel_RRA.m by Martin M. Andreasen, Jan 2016 + +@#include "AFVRR_common.inc" + +%-------------------------------------------------------------------------- +% Parameter calibration taken from RunGMM_standardModel_RRA.m +%-------------------------------------------------------------------------- +% fixed parameters +INHABIT = 1; +PHI1 = 4; +PHI4 = 1; +KAPAone = 0; +DELTA = 0.025; +THETA = 0.36; +ETA = 6; +CHI = 0; +CONSxhr40 = 0; +BETTAxhr = 0; +BETTAxhr40= 0; +RHOD = 0; +GAMA = 0.9999; +CONSxhr20 = 0; + +% estimated parameters +BETTA = 0.999544966118000; +B = 0.668859504661000; +H = 0.342483445196000; +PHI2 = 0.997924964981000; +RRA = 662.7953149595370; +KAPAtwo = 5.516226495551000; +ALFA = 0.809462321180000; +RHOR = 0.643873352513000; +BETTAPAI = 1.270087844103000; +BETTAY = 0.031812764291000; +MYYPS = 1.001189151180000; +MYZ = 1.005286347928000; +RHOA = 0.743239127127000; +RHOG = 0.793929380230000; +PAI = 1.012163659169000; +GoY = 0.206594858866000; +STDA = 0.016586292524000; +STDG = 0.041220613851000; +STDD = 0.013534473123000; + +% endogenous parameters set via steady state, no need to initialize +%PHIzero = ; +%AA = ; +%PHI3 = ; +%negVf = ; + +model_diagnostics; +% Model diagnostics show that some parameters are endogenously determined +% via the steady state, so we run steady to calibrate all parameters +steady; +model_diagnostics; +% Now all parameters are determined + +resid; +check; + +%-------------------------------------------------------------------------- +% Shock distribution +%-------------------------------------------------------------------------- +shocks; +var eps_a = STDA^2; +var eps_d = STDD^2; +var eps_g = STDG^2; +end; + +%-------------------------------------------------------------------------- +% Estimated Params block - these parameters will be estimated, we +% initialize at calibrated values +%-------------------------------------------------------------------------- +estimated_params; +BETTA; +B; +H; +PHI2; +RRA; +KAPAtwo; +ALFA; +RHOR; +BETTAPAI; +BETTAY; +MYYPS; +MYZ; +RHOA; +RHOG; +PAI; +GoY; +stderr eps_a; +stderr eps_g; +stderr eps_d; +end; + +estimated_params_init(use_calibration); +end; + +%-------------------------------------------------------------------------- +% Compare whether toolbox yields equivalent moments at second order +%-------------------------------------------------------------------------- +% Note that we compare results for orderApp=1|2 and not for orderApp=3, because +% there is a small error in the replication files of the original article in the +% computation of the covariance matrix of the extended innovations vector. +% The authors have been contacted, fixed it, and report that the results +% change only slightly at orderApp=3 to what they report in the paper. At +% orderApp=2 all is correct and so the following part tests whether we get +% the same model moments at the calibrated parameters (we do not optimize). +% We compare it to the replication file RunGMM_standardModel_RRA.m with the +% following settings: orderApp=1|2, seOn=0, q_lag=10, weighting=1; +% scaled=0; optimizer=0; estimator=1; momentSet=2; +% +% Output of the replication files for orderApp=1 +AndreasenEtAl.Q1 = 23893.072; +AndreasenEtAl.moments1 =[ % note that we reshuffeled to be compatible with our matched moments block + {[ 1]} {'Ex' } {'Gr_C '} {' ' } {'0.024388' } {'0.023764' } + {[ 2]} {'Ex' } {'Gr_I '} {' ' } {'0.031046' } {'0.028517' } + {[ 3]} {'Ex' } {'Infl ' } {' ' } {'0.03757' } {'0.048361' } + {[ 4]} {'Ex' } {'r1 ' } {' ' } {'0.056048' } {'0.073945' } + {[ 5]} {'Ex' } {'r40 ' } {' ' } {'0.069929' } {'0.073945' } + {[ 6]} {'Ex' } {'xhr40 '} {' ' } {'0.017237' } {'0' } + {[ 7]} {'Ex' } {'GoY '} {' ' } {'-1.5745' } {'-1.577' } + {[ 8]} {'Ex' } {'hours '} {' ' } {'-0.043353' } {'-0.042861' } + {[ 9]} {'Exx' } {'Gr_C '} {'Gr_C '} {'0.0013159' } {'0.0011816' } + {[17]} {'Exx' } {'Gr_C '} {'Gr_I '} {'0.0021789' } {'0.0016052' } + {[18]} {'Exx' } {'Gr_C '} {'Infl ' } {'0.00067495' } {'0.00090947' } + {[19]} {'Exx' } {'Gr_C '} {'r1 ' } {'0.0011655' } {'0.0016016' } + {[20]} {'Exx' } {'Gr_C '} {'r40 ' } {'0.0015906' } {'0.0017076' } + {[21]} {'Exx' } {'Gr_C '} {'xhr40 '} {'0.0020911' } {'0.0013997' } + {[10]} {'Exx' } {'Gr_I '} {'Gr_I '} {'0.0089104' } {'0.0055317' } + {[22]} {'Exx' } {'Gr_I '} {'Infl ' } {'0.00063139' } {'0.00050106' } + {[23]} {'Exx' } {'Gr_I '} {'r1 ' } {'0.0011031' } {'0.0018178' } + {[24]} {'Exx' } {'Gr_I '} {'r40 ' } {'0.0018445' } {'0.0020186' } + {[25]} {'Exx' } {'Gr_I '} {'xhr40 '} {'0.00095556' } {'0.0064471' } + {[11]} {'Exx' } {'Infl ' } {'Infl ' } {'0.0020268' } {'0.0030519' } + {[26]} {'Exx' } {'Infl ' } {'r1 ' } {'0.0025263' } {'0.0042181' } + {[27]} {'Exx' } {'Infl ' } {'r40 ' } {'0.0029126' } {'0.0039217' } + {[28]} {'Exx' } {'Infl ' } {'xhr40 '} {'-0.00077101'} {'-0.0019975' } + {[12]} {'Exx' } {'r1 ' } {'r1 ' } {'0.0038708' } {'0.0061403' } + {[29]} {'Exx' } {'r1 ' } {'r40 ' } {'0.0044773' } {'0.0058343' } + {[30]} {'Exx' } {'r1 ' } {'xhr40 '} {'-0.00048202'} {'-0.00089501'} + {[13]} {'Exx' } {'r40 ' } {'r40 ' } {'0.0054664' } {'0.0056883' } + {[31]} {'Exx' } {'r40 ' } {'xhr40 '} {'0.00053864' } {'-0.00041184'} + {[14]} {'Exx' } {'xhr40 '} {'xhr40 '} {'0.053097' } {'0.016255' } + {[15]} {'Exx' } {'GoY '} {'GoY '} {'2.4863' } {'2.4919' } + {[16]} {'Exx' } {'hours '} {'hours '} {'0.0018799' } {'0.0018384' } + {[32]} {'Exx1'} {'Gr_C '} {'Gr_C '} {'0.00077917' } {'0.00065543' } + {[33]} {'Exx1'} {'Gr_I '} {'Gr_I '} {'0.0050104' } {'0.0033626' } + {[34]} {'Exx1'} {'Infl ' } {'Infl ' } {'0.0019503' } {'0.0029033' } + {[35]} {'Exx1'} {'r1 ' } {'r1 ' } {'0.0038509' } {'0.006112' } + {[36]} {'Exx1'} {'r40 ' } {'r40 ' } {'0.0054699' } {'0.005683' } + {[37]} {'Exx1'} {'xhr40 '} {'xhr40 '} {'-0.00098295'} {'3.3307e-16' } + {[38]} {'Exx1'} {'GoY '} {'GoY '} {'2.4868' } {'2.4912' } + {[39]} {'Exx1'} {'hours '} {'hours '} {'0.0018799' } {'0.0018378' } +]; + +% Output of the replication files for orderApp=2 +AndreasenEtAl.Q2 = 65.8269; +AndreasenEtAl.moments2 =[ % note that we reshuffeled to be compatible with our matched moments block + {[ 1]} {'Ex' } {'Gr_C '} {' ' } {'0.024388' } {'0.023764' } + {[ 2]} {'Ex' } {'Gr_I '} {' ' } {'0.031046' } {'0.028517' } + {[ 3]} {'Ex' } {'Infl ' } {' ' } {'0.03757' } {'0.034882' } + {[ 4]} {'Ex' } {'r1 ' } {' ' } {'0.056048' } {'0.056542' } + {[ 5]} {'Ex' } {'r40 ' } {' ' } {'0.069929' } {'0.070145' } + {[ 6]} {'Ex' } {'xhr40 '} {' ' } {'0.017237' } {'0.020825' } + {[ 7]} {'Ex' } {'GoY '} {' ' } {'-1.5745' } {'-1.5748' } + {[ 8]} {'Ex' } {'hours '} {' ' } {'-0.043353' } {'-0.04335' } + {[ 9]} {'Exx' } {'Gr_C '} {'Gr_C '} {'0.0013159' } {'0.001205' } + {[17]} {'Exx' } {'Gr_C '} {'Gr_I '} {'0.0021789' } {'0.0016067' } + {[18]} {'Exx' } {'Gr_C '} {'Infl ' } {'0.00067495' } {'0.00059406'} + {[19]} {'Exx' } {'Gr_C '} {'r1 ' } {'0.0011655' } {'0.0011949' } + {[20]} {'Exx' } {'Gr_C '} {'r40 ' } {'0.0015906' } {'0.0016104' } + {[21]} {'Exx' } {'Gr_C '} {'xhr40 '} {'0.0020911' } {'0.0020245' } + {[10]} {'Exx' } {'Gr_I '} {'Gr_I '} {'0.0089104' } {'0.0060254' } + {[22]} {'Exx' } {'Gr_I '} {'Infl ' } {'0.00063139' } {'8.3563e-05'} + {[23]} {'Exx' } {'Gr_I '} {'r1 ' } {'0.0011031' } {'0.0013176' } + {[24]} {'Exx' } {'Gr_I '} {'r40 ' } {'0.0018445' } {'0.0019042' } + {[25]} {'Exx' } {'Gr_I '} {'xhr40 '} {'0.00095556' } {'0.0064261' } + {[11]} {'Exx' } {'Infl ' } {'Infl ' } {'0.0020268' } {'0.0020735' } + {[26]} {'Exx' } {'Infl ' } {'r1 ' } {'0.0025263' } {'0.0027621' } + {[27]} {'Exx' } {'Infl ' } {'r40 ' } {'0.0029126' } {'0.0029257' } + {[28]} {'Exx' } {'Infl ' } {'xhr40 '} {'-0.00077101'} {'-0.0012165'} + {[12]} {'Exx' } {'r1 ' } {'r1 ' } {'0.0038708' } {'0.0040235' } + {[29]} {'Exx' } {'r1 ' } {'r40 ' } {'0.0044773' } {'0.0044702' } + {[30]} {'Exx' } {'r1 ' } {'xhr40 '} {'-0.00048202'} {'0.00030542'} + {[13]} {'Exx' } {'r40 ' } {'r40 ' } {'0.0054664' } {'0.0052718' } + {[31]} {'Exx' } {'r40 ' } {'xhr40 '} {'0.00053864' } {'0.0010045' } + {[14]} {'Exx' } {'xhr40 '} {'xhr40 '} {'0.053097' } {'0.018416' } + {[15]} {'Exx' } {'GoY '} {'GoY '} {'2.4863' } {'2.4853' } + {[16]} {'Exx' } {'hours '} {'hours '} {'0.0018799' } {'0.0018806' } + {[32]} {'Exx1'} {'Gr_C '} {'Gr_C '} {'0.00077917' } {'0.00067309'} + {[33]} {'Exx1'} {'Gr_I '} {'Gr_I '} {'0.0050104' } {'0.0033293' } + {[34]} {'Exx1'} {'Infl ' } {'Infl ' } {'0.0019503' } {'0.0019223' } + {[35]} {'Exx1'} {'r1 ' } {'r1 ' } {'0.0038509' } {'0.0039949' } + {[36]} {'Exx1'} {'r40 ' } {'r40 ' } {'0.0054699' } {'0.0052659' } + {[37]} {'Exx1'} {'xhr40 '} {'xhr40 '} {'-0.00098295'} {'0.0004337' } + {[38]} {'Exx1'} {'GoY '} {'GoY '} {'2.4868' } {'2.4846' } + {[39]} {'Exx1'} {'hours '} {'hours '} {'0.0018799' } {'0.00188' } +]; + +@#for orderApp in 1:2 + +method_of_moments( + mom_method = GMM % method of moments method; possible values: GMM|SMM + , datafile = 'AFVRR_data.mat' % name of filename with data + , bartlett_kernel_lag = 10 % bandwith in optimal weighting matrix + , order = @{orderApp} % order of Taylor approximation in perturbation + , pruning % use pruned state space system at higher-order + % , verbose % display and store intermediate estimation results + , weighting_matrix = ['DIAGONAL'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename + % , TeX % print TeX tables and graphics + % Optimization options that can be set by the user in the mod file, otherwise default values are provided + %, huge_number=1D10 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons + , mode_compute = 0 % specifies the optimizer for minimization of moments distance, note that by default there is a new optimizer + , optim = ('TolFun', 1e-6 + ,'TolX', 1e-6 + ,'MaxIter', 3000 + ,'MaxFunEvals', 1D6 + ,'UseParallel' , 1 + %,'Jacobian' , 'on' + ) % a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute + %, silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + %, analytic_standard_errors + , se_tolx=1e-10 +); + +% Check results + +fprintf('****************************************************************\n') +fprintf('Compare Results for perturbation order @{orderApp}\n') +fprintf('****************************************************************\n') +dev_Q = AndreasenEtAl.Q@{orderApp} - oo_.mom.Q; +dev_datamoments = str2double(AndreasenEtAl.moments@{orderApp}(:,5)) - oo_.mom.data_moments; +dev_modelmoments = str2double(AndreasenEtAl.moments@{orderApp}(:,6)) - oo_.mom.model_moments; + +table([AndreasenEtAl.Q@{orderApp} ; str2double(AndreasenEtAl.moments@{orderApp}(:,5)) ; str2double(AndreasenEtAl.moments@{orderApp}(:,6))],... + [oo_.mom.Q ; oo_.mom.data_moments ; oo_.mom.model_moments ],... + [dev_Q ; dev_datamoments ; dev_modelmoments ],... + 'VariableNames', {'Andreasen et al', 'Dynare', 'dev'},... + 'RowNames', ['Q'; strcat('Data_', M_.matched_moments(:,4)); strcat('Model_', M_.matched_moments(:,4))]) + +if norm(dev_modelmoments)> 1e-4 + error('Something wrong in the computation of moments at order @{orderApp}') +end + +@#endfor + +%-------------------------------------------------------------------------- +% Replicate estimation at orderApp=3 +%-------------------------------------------------------------------------- +@#ifdef DoEstimation +method_of_moments( + mom_method = GMM % method of moments method; possible values: GMM|SMM + , datafile = 'AFVRR_data.mat' % name of filename with data + , bartlett_kernel_lag = 10 % bandwith in optimal weighting matrix + , order = 3 % order of Taylor approximation in perturbation + , pruning % use pruned state space system at higher-order + % , verbose % display and store intermediate estimation results + , weighting_matrix = ['DIAGONAL', 'OPTIMAL'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename + % , TeX % print TeX tables and graphics + % Optimization options that can be set by the user in the mod file, otherwise default values are provided + %, huge_number=1D10 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons + , mode_compute = 13 % specifies the optimizer for minimization of moments distance, note that by default there is a new optimizer + , additional_optimizer_steps = [13] + , optim = ('TolFun', 1e-6 + ,'TolX', 1e-6 + ,'MaxIter', 3000 + ,'MaxFunEvals', 1D6 + ,'UseParallel' , 1 + %,'Jacobian' , 'on' + ) % a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute + %, silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + %, analytic_standard_errors + , se_tolx=1e-10 +); +@#endif \ No newline at end of file diff --git a/tests/estimation/method_of_moments/AFVRR/AFVRR_MFB.mod b/tests/estimation/method_of_moments/AFVRR/AFVRR_MFB.mod new file mode 100644 index 000000000..450739ad3 --- /dev/null +++ b/tests/estimation/method_of_moments/AFVRR/AFVRR_MFB.mod @@ -0,0 +1,300 @@ +% DSGE model based on replication files of +% Andreasen, Fernandez-Villaverde, Rubio-Ramirez (2018), The Pruned State-Space System for Non-Linear DSGE Models: Theory and Empirical Applications, Review of Economic Studies, 85, p. 1-49 +% Adapted for Dynare by Willi Mutschler (@wmutschl, willi@mutschler.eu), Jan 2021 +% ========================================================================= +% Copyright (C) 2021 Dynare Team +% +% This file is part of Dynare. +% +% Dynare is free software: you can redistribute it and/or modify +% it under the terms of the GNU General Public License as published by +% the Free Software Foundation, either version 3 of the License, or +% (at your option) any later version. +% +% Dynare is distributed in the hope that it will be useful, +% but WITHOUT ANY WARRANTY; without even the implied warranty of +% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +% GNU General Public License for more details. +% +% You should have received a copy of the GNU General Public License +% along with Dynare. If not, see . +% ========================================================================= + +% This is the model with Feedback M_FB +% Original code RunGMM_Feedback_estim_RRA.m by Martin M. Andreasen, Jan 2016 + +@#include "AFVRR_common.inc" + +%-------------------------------------------------------------------------- +% Parameter calibration taken from RunGMM_Feedback_estim_RRA.m +%-------------------------------------------------------------------------- +% fixed parameters +INHABIT = 1; +PHI1 = 4; +PHI4 = 1; +KAPAone = 0; +DELTA = 0.025; +THETA = 0.36; +ETA = 6; +CHI = 0; +BETTAxhr = 0; +BETTAxhr40= 0; +RHOD = 0; +GAMA = 0.9999; +CONSxhr20 = 0; + +% estimated parameters +BETTA = 0.997007023687000; +B = 0.692501768577000; +H = 0.339214495653000; +PHI2 = 0.688555040951000; +RRA = 24.346514272871001; +KAPAtwo = 10.018421876923000; +ALFA = 0.792507553312000; +RHOR = 0.849194030384000; +BETTAPAI = 2.060579322980000; +BETTAY = 0.220573712342000; +MYYPS = 1.001016690133000; +MYZ = 1.005356313981000; +RHOA = 0.784141391843000; +RHOG = 0.816924540497000; +PAI = 1.011924196487000; +CONSxhr40 = 0.878774662208000; +GoY = 0.207110300602000; +STDA = 0.013024450606000; +STDG = 0.051049871928000; +STDD = 0.008877423780000; + +% endogenous parameters set via steady state, no need to initialize +%PHIzero = ; +%AA = ; +%PHI3 = ; +%negVf = ; + +model_diagnostics; +% Model diagnostics show that some parameters are endogenously determined +% via the steady state, so we run steady to calibrate all parameters +steady; +model_diagnostics; +% Now all parameters are determined + +resid; +check; + +%-------------------------------------------------------------------------- +% Shock distribution +%-------------------------------------------------------------------------- +shocks; +var eps_a = STDA^2; +var eps_d = STDD^2; +var eps_g = STDG^2; +end; + +%-------------------------------------------------------------------------- +% Estimated Params block - these parameters will be estimated, we +% initialize at calibrated values +%-------------------------------------------------------------------------- +estimated_params; +BETTA; +B; +H; +PHI2; +RRA; +KAPAtwo; +ALFA; +RHOR; +BETTAPAI; +BETTAY; +MYYPS; +MYZ; +RHOA; +RHOG; +PAI; +CONSxhr40; +GoY; +stderr eps_a; +stderr eps_g; +stderr eps_d; +end; + +estimated_params_init(use_calibration); +end; + +%-------------------------------------------------------------------------- +% Compare whether toolbox yields equivalent moments at second order +%-------------------------------------------------------------------------- +% Note that we compare results for orderApp=1|2 and not for orderApp=3, because +% there is a small error in the replication files of the original article in the +% computation of the covariance matrix of the extended innovations vector. +% The authors have been contacted, fixed it, and report that the results +% change only slightly at orderApp=3 to what they report in the paper. At +% orderApp=2 all is correct and so the following part tests whether we get +% the same model moments at the calibrated parameters (we do not optimize). +% We compare it to the replication file RunGMM_Feedback_estim_RRA.m with the +% following settings: orderApp=1|2, seOn=0, q_lag=10, weighting=1; +% scaled=0; optimizer=0; estimator=1; momentSet=2; +% +% Output of the replication files for orderApp=1 +AndreasenEtAl.Q1 = 201778.9697; +AndreasenEtAl.moments1 =[ % note that we reshuffeled to be compatible with our matched moments block + {[ 1]} {'Ex' } {'Gr_C '} {' ' } {'0.024388' } {'0.023654' } + {[ 2]} {'Ex' } {'Gr_I '} {' ' } {'0.031046' } {'0.027719' } + {[ 3]} {'Ex' } {'Infl ' } {' ' } {'0.03757' } {'0.047415' } + {[ 4]} {'Ex' } {'r1 ' } {' ' } {'0.056048' } {'0.083059' } + {[ 5]} {'Ex' } {'r40 ' } {' ' } {'0.069929' } {'0.083059' } + {[ 6]} {'Ex' } {'xhr40 '} {' ' } {'0.017237' } {'0' } + {[ 7]} {'Ex' } {'GoY '} {' ' } {'-1.5745' } {'-1.5745' } + {[ 8]} {'Ex' } {'hours '} {' ' } {'-0.043353' } {'-0.043245' } + {[ 9]} {'Exx' } {'Gr_C '} {'Gr_C '} {'0.0013159' } {'0.0012253' } + {[17]} {'Exx' } {'Gr_C '} {'Gr_I '} {'0.0021789' } {'0.0015117' } + {[18]} {'Exx' } {'Gr_C '} {'Infl ' } {'0.00067495' } {'0.00080078' } + {[19]} {'Exx' } {'Gr_C '} {'r1 ' } {'0.0011655' } {'0.00182' } + {[20]} {'Exx' } {'Gr_C '} {'r40 ' } {'0.0015906' } {'0.001913' } + {[21]} {'Exx' } {'Gr_C '} {'xhr40 '} {'0.0020911' } {'0.0016326' } + {[10]} {'Exx' } {'Gr_I '} {'Gr_I '} {'0.0089104' } {'0.0040112' } + {[22]} {'Exx' } {'Gr_I '} {'Infl ' } {'0.00063139' } {'0.00060604' } + {[23]} {'Exx' } {'Gr_I '} {'r1 ' } {'0.0011031' } {'0.0021426' } + {[24]} {'Exx' } {'Gr_I '} {'r40 ' } {'0.0018445' } {'0.0022348' } + {[25]} {'Exx' } {'Gr_I '} {'xhr40 '} {'0.00095556' } {'0.0039852' } + {[11]} {'Exx' } {'Infl ' } {'Infl ' } {'0.0020268' } {'0.0030058' } + {[26]} {'Exx' } {'Infl ' } {'r1 ' } {'0.0025263' } {'0.0044951' } + {[27]} {'Exx' } {'Infl ' } {'r40 ' } {'0.0029126' } {'0.0042225' } + {[28]} {'Exx' } {'Infl ' } {'xhr40 '} {'-0.00077101'} {'-0.0021222' } + {[12]} {'Exx' } {'r1 ' } {'r1 ' } {'0.0038708' } {'0.0074776' } + {[29]} {'Exx' } {'r1 ' } {'r40 ' } {'0.0044773' } {'0.0071906' } + {[30]} {'Exx' } {'r1 ' } {'xhr40 '} {'-0.00048202'} {'-0.0006736' } + {[13]} {'Exx' } {'r40 ' } {'r40 ' } {'0.0054664' } {'0.0070599' } + {[31]} {'Exx' } {'r40 ' } {'xhr40 '} {'0.00053864' } {'-0.00036735'} + {[14]} {'Exx' } {'xhr40 '} {'xhr40 '} {'0.053097' } {'0.014516' } + {[15]} {'Exx' } {'GoY '} {'GoY '} {'2.4863' } {'2.4866' } + {[16]} {'Exx' } {'hours '} {'hours '} {'0.0018799' } {'0.0018713' } + {[32]} {'Exx1'} {'Gr_C '} {'Gr_C '} {'0.00077917' } {'0.00076856' } + {[33]} {'Exx1'} {'Gr_I '} {'Gr_I '} {'0.0050104' } {'0.002163' } + {[34]} {'Exx1'} {'Infl ' } {'Infl ' } {'0.0019503' } {'0.0028078' } + {[35]} {'Exx1'} {'r1 ' } {'r1 ' } {'0.0038509' } {'0.0074583' } + {[36]} {'Exx1'} {'r40 ' } {'r40 ' } {'0.0054699' } {'0.0070551' } + {[37]} {'Exx1'} {'xhr40 '} {'xhr40 '} {'-0.00098295'} {'7.2164e-16' } + {[38]} {'Exx1'} {'GoY '} {'GoY '} {'2.4868' } {'2.4856' } + {[39]} {'Exx1'} {'hours '} {'hours '} {'0.0018799' } {'0.0018708' } +]; + +% Output of the replication files for orderApp=2 +AndreasenEtAl.Q2 = 59.3323; +AndreasenEtAl.moments2 =[ % note that we reshuffeled to be compatible with our matched moments block + {[ 1]} {'Ex' } {'Gr_C '} {' ' } {'0.024388' } {'0.023654' } + {[ 2]} {'Ex' } {'Gr_I '} {' ' } {'0.031046' } {'0.027719' } + {[ 3]} {'Ex' } {'Infl ' } {' ' } {'0.03757' } {'0.034565' } + {[ 4]} {'Ex' } {'r1 ' } {' ' } {'0.056048' } {'0.056419' } + {[ 5]} {'Ex' } {'r40 ' } {' ' } {'0.069929' } {'0.07087' } + {[ 6]} {'Ex' } {'xhr40 '} {' ' } {'0.017237' } {'0.01517' } + {[ 7]} {'Ex' } {'GoY '} {' ' } {'-1.5745' } {'-1.5743' } + {[ 8]} {'Ex' } {'hours '} {' ' } {'-0.043353' } {'-0.043352' } + {[ 9]} {'Exx' } {'Gr_C '} {'Gr_C '} {'0.0013159' } {'0.0012464' } + {[17]} {'Exx' } {'Gr_C '} {'Gr_I '} {'0.0021789' } {'0.0015247' } + {[18]} {'Exx' } {'Gr_C '} {'Infl ' } {'0.00067495' } {'0.0004867' } + {[19]} {'Exx' } {'Gr_C '} {'r1 ' } {'0.0011655' } {'0.0011867' } + {[20]} {'Exx' } {'Gr_C '} {'r40 ' } {'0.0015906' } {'0.0016146' } + {[21]} {'Exx' } {'Gr_C '} {'xhr40 '} {'0.0020911' } {'0.0021395' } + {[10]} {'Exx' } {'Gr_I '} {'Gr_I '} {'0.0089104' } {'0.0043272' } + {[22]} {'Exx' } {'Gr_I '} {'Infl ' } {'0.00063139' } {'0.00021752'} + {[23]} {'Exx' } {'Gr_I '} {'r1 ' } {'0.0011031' } {'0.0013919' } + {[24]} {'Exx' } {'Gr_I '} {'r40 ' } {'0.0018445' } {'0.0018899' } + {[25]} {'Exx' } {'Gr_I '} {'xhr40 '} {'0.00095556' } {'0.0037854' } + {[11]} {'Exx' } {'Infl ' } {'Infl ' } {'0.0020268' } {'0.0021043' } + {[26]} {'Exx' } {'Infl ' } {'r1 ' } {'0.0025263' } {'0.0026571' } + {[27]} {'Exx' } {'Infl ' } {'r40 ' } {'0.0029126' } {'0.0028566' } + {[28]} {'Exx' } {'Infl ' } {'xhr40 '} {'-0.00077101'} {'-0.0016279'} + {[12]} {'Exx' } {'r1 ' } {'r1 ' } {'0.0038708' } {'0.0039136' } + {[29]} {'Exx' } {'r1 ' } {'r40 ' } {'0.0044773' } {'0.0044118' } + {[30]} {'Exx' } {'r1 ' } {'xhr40 '} {'-0.00048202'} {'0.00016791'} + {[13]} {'Exx' } {'r40 ' } {'r40 ' } {'0.0054664' } {'0.0052851' } + {[31]} {'Exx' } {'r40 ' } {'xhr40 '} {'0.00053864' } {'0.00062143'} + {[14]} {'Exx' } {'xhr40 '} {'xhr40 '} {'0.053097' } {'0.018126' } + {[15]} {'Exx' } {'GoY '} {'GoY '} {'2.4863' } {'2.4863' } + {[16]} {'Exx' } {'hours '} {'hours '} {'0.0018799' } {'0.0018806' } + {[32]} {'Exx1'} {'Gr_C '} {'Gr_C '} {'0.00077917' } {'0.00078586'} + {[33]} {'Exx1'} {'Gr_I '} {'Gr_I '} {'0.0050104' } {'0.0021519' } + {[34]} {'Exx1'} {'Infl ' } {'Infl ' } {'0.0019503' } {'0.0019046' } + {[35]} {'Exx1'} {'r1 ' } {'r1 ' } {'0.0038509' } {'0.0038939' } + {[36]} {'Exx1'} {'r40 ' } {'r40 ' } {'0.0054699' } {'0.0052792' } + {[37]} {'Exx1'} {'xhr40 '} {'xhr40 '} {'-0.00098295'} {'0.00023012'} + {[38]} {'Exx1'} {'GoY '} {'GoY '} {'2.4868' } {'2.4852' } + {[39]} {'Exx1'} {'hours '} {'hours '} {'0.0018799' } {'0.0018801' } +]; + +@#for orderApp in 1:2 + +method_of_moments( + mom_method = GMM % method of moments method; possible values: GMM|SMM + , datafile = 'AFVRR_data.mat' % name of filename with data + , bartlett_kernel_lag = 10 % bandwith in optimal weighting matrix + , order = @{orderApp} % order of Taylor approximation in perturbation + , pruning % use pruned state space system at higher-order + % , verbose % display and store intermediate estimation results + , weighting_matrix = ['DIAGONAL'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename + % , TeX % print TeX tables and graphics + % Optimization options that can be set by the user in the mod file, otherwise default values are provided + %, huge_number=1D10 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons + , mode_compute = 0 % specifies the optimizer for minimization of moments distance, note that by default there is a new optimizer + , optim = ('TolFun', 1e-6 + ,'TolX', 1e-6 + ,'MaxIter', 3000 + ,'MaxFunEvals', 1D6 + ,'UseParallel' , 1 + %,'Jacobian' , 'on' + ) % a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute + %, silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + %, analytic_standard_errors + , se_tolx=1e-10 +); + +% Check results + +fprintf('****************************************************************\n') +fprintf('Compare Results for perturbation order @{orderApp}\n') +fprintf('****************************************************************\n') +dev_Q = AndreasenEtAl.Q@{orderApp} - oo_.mom.Q; +dev_datamoments = str2double(AndreasenEtAl.moments@{orderApp}(:,5)) - oo_.mom.data_moments; +dev_modelmoments = str2double(AndreasenEtAl.moments@{orderApp}(:,6)) - oo_.mom.model_moments; + +table([AndreasenEtAl.Q@{orderApp} ; str2double(AndreasenEtAl.moments@{orderApp}(:,5)) ; str2double(AndreasenEtAl.moments@{orderApp}(:,6))],... + [oo_.mom.Q ; oo_.mom.data_moments ; oo_.mom.model_moments ],... + [dev_Q ; dev_datamoments ; dev_modelmoments ],... + 'VariableNames', {'Andreasen et al', 'Dynare', 'dev'},... + 'RowNames', ['Q'; strcat('Data_', M_.matched_moments(:,4)); strcat('Model_', M_.matched_moments(:,4))]) + +if norm(dev_modelmoments)> 1e-4 + warning('Something wrong in the computation of moments at order @{orderApp}') +end + +@#endfor + +%-------------------------------------------------------------------------- +% Replicate estimation at orderApp=3 +%-------------------------------------------------------------------------- +@#ifdef DoEstimation +method_of_moments( + mom_method = GMM % method of moments method; possible values: GMM|SMM + , datafile = 'AFVRR_data.mat' % name of filename with data + , bartlett_kernel_lag = 10 % bandwith in optimal weighting matrix + , order = 3 % order of Taylor approximation in perturbation + , pruning % use pruned state space system at higher-order + % , verbose % display and store intermediate estimation results + , weighting_matrix = ['DIAGONAL', 'Optimal'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename + % , TeX % print TeX tables and graphics + % Optimization options that can be set by the user in the mod file, otherwise default values are provided + %, huge_number=1D10 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons + , mode_compute = 13 % specifies the optimizer for minimization of moments distance, note that by default there is a new optimizer + , additional_optimizer_steps = [13] + , optim = ('TolFun', 1e-6 + ,'TolX', 1e-6 + ,'MaxIter', 3000 + ,'MaxFunEvals', 1D6 + ,'UseParallel' , 1 + %,'Jacobian' , 'on' + ) % a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute + %, silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + %, analytic_standard_errors + , se_tolx=1e-10 +); +@#endif \ No newline at end of file diff --git a/tests/estimation/method_of_moments/AFVRR/AFVRR_MFB_RRA.mod b/tests/estimation/method_of_moments/AFVRR/AFVRR_MFB_RRA.mod new file mode 100644 index 000000000..9c069d3a3 --- /dev/null +++ b/tests/estimation/method_of_moments/AFVRR/AFVRR_MFB_RRA.mod @@ -0,0 +1,299 @@ +% DSGE model based on replication files of +% Andreasen, Fernandez-Villaverde, Rubio-Ramirez (2018), The Pruned State-Space System for Non-Linear DSGE Models: Theory and Empirical Applications, Review of Economic Studies, 85, p. 1-49 +% Adapted for Dynare by Willi Mutschler (@wmutschl, willi@mutschler.eu), Jan 2021 +% ========================================================================= +% Copyright (C) 2021 Dynare Team +% +% This file is part of Dynare. +% +% Dynare is free software: you can redistribute it and/or modify +% it under the terms of the GNU General Public License as published by +% the Free Software Foundation, either version 3 of the License, or +% (at your option) any later version. +% +% Dynare is distributed in the hope that it will be useful, +% but WITHOUT ANY WARRANTY; without even the implied warranty of +% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +% GNU General Public License for more details. +% +% You should have received a copy of the GNU General Public License +% along with Dynare. If not, see . +% ========================================================================= + +% This is the model with feedback and calibrated RRA +% Original code RunGMM_Feedback_estim_RRA_5.m by Martin M. Andreasen, Jan 2016 + +@#include "AFVRR_common.inc" + +%-------------------------------------------------------------------------- +% Parameter calibration taken from RunGMM_Feedback_estim_RRA_5.m +%-------------------------------------------------------------------------- +% fixed parameters +INHABIT = 1; +PHI1 = 4; +PHI4 = 1; +KAPAone = 0; +DELTA = 0.025; +THETA = 0.36; +ETA = 6; +CHI = 0; +BETTAxhr = 0; +BETTAxhr40= 0; +RHOD = 0; +GAMA = 0.9999; +CONSxhr20 = 0; +RRA = 5; + +% estimated parameters +BETTA = 0.996850651147000; +B = 0.684201133923000; +H = 0.338754441432000; +PHI2 = 0.738293581320000; +KAPAtwo = 11.664785970704999; +ALFA = 0.831836572237000; +RHOR = 0.772754520116000; +BETTAPAI = 3.020381242896000; +BETTAY = 0.288367683973000; +MYYPS = 1.000911709188000; +MYZ = 1.005433723022000; +RHOA = 0.749465413198000; +RHOG = 0.847225569814000; +PAI = 1.010428794858000; +CONSxhr40 = 0.992863217133000; +GoY = 0.207099399789000; +STDA = 0.015621059978000; +STDG = 0.047539390956000; +STDD = 0.008623441943000; + +% endogenous parameters set via steady state, no need to initialize +%PHIzero = ; +%AA = ; +%PHI3 = ; +%negVf = ; + +model_diagnostics; +% Model diagnostics show that some parameters are endogenously determined +% via the steady state, so we run steady to calibrate all parameters +steady; +model_diagnostics; +% Now all parameters are determined + +resid; +check; + +%-------------------------------------------------------------------------- +% Shock distribution +%-------------------------------------------------------------------------- +shocks; +var eps_a = STDA^2; +var eps_d = STDD^2; +var eps_g = STDG^2; +end; + +%-------------------------------------------------------------------------- +% Estimated Params block - these parameters will be estimated, we +% initialize at calibrated values +%-------------------------------------------------------------------------- +estimated_params; +BETTA; +B; +H; +PHI2; +KAPAtwo; +ALFA; +RHOR; +BETTAPAI; +BETTAY; +MYYPS; +MYZ; +RHOA; +RHOG; +PAI; +CONSxhr40; +GoY; +stderr eps_a; +stderr eps_g; +stderr eps_d; +end; + +estimated_params_init(use_calibration); +end; + +%-------------------------------------------------------------------------- +% Compare whether toolbox yields equivalent moments at second order +%-------------------------------------------------------------------------- +% Note that we compare results for orderApp=1|2 and not for orderApp=3, because +% there is a small error in the replication files of the original article in the +% computation of the covariance matrix of the extended innovations vector. +% The authors have been contacted, fixed it, and report that the results +% change only slightly at orderApp=3 to what they report in the paper. At +% orderApp=2 all is correct and so the following part tests whether we get +% the same model moments at the calibrated parameters (we do not optimize). +% We compare it to the replication file RunGMM_Feedback_estim_RRA.m with the +% following settings: orderApp=1|2, seOn=1, q_lag=10, weighting=1+1; +% scaled=0; optimizer=0; estimator=1; momentSet=2; +% +% Output of the replication files for orderApp=1 +AndreasenEtAl.Q1 = 60275.3715; +AndreasenEtAl.moments1 =[ % note that we reshuffeled to be compatible with our matched moments block + {[ 1]} {'Ex' } {'Gr_C '} {' ' } {'0.024388' } {'0.023726' } + {[ 2]} {'Ex' } {'Gr_I '} {' ' } {'0.031046' } {'0.027372' } + {[ 3]} {'Ex' } {'Infl ' } {' ' } {'0.03757' } {'0.041499' } + {[ 4]} {'Ex' } {'r1 ' } {' ' } {'0.056048' } {'0.077843' } + {[ 5]} {'Ex' } {'r40 ' } {' ' } {'0.069929' } {'0.077843' } + {[ 6]} {'Ex' } {'xhr40 '} {' ' } {'0.017237' } {'0' } + {[ 7]} {'Ex' } {'GoY '} {' ' } {'-1.5745' } {'-1.5746' } + {[ 8]} {'Ex' } {'hours '} {' ' } {'-0.043353' } {'-0.043299' } + {[ 9]} {'Exx' } {'Gr_C '} {'Gr_C '} {'0.0013159' } {'0.0012763' } + {[17]} {'Exx' } {'Gr_C '} {'Gr_I '} {'0.0021789' } {'0.0017759' } + {[18]} {'Exx' } {'Gr_C '} {'Infl ' } {'0.00067495' } {'0.00077354' } + {[19]} {'Exx' } {'Gr_C '} {'r1 ' } {'0.0011655' } {'0.0016538' } + {[20]} {'Exx' } {'Gr_C '} {'r40 ' } {'0.0015906' } {'0.0017949' } + {[21]} {'Exx' } {'Gr_C '} {'xhr40 '} {'0.0020911' } {'0.0017847' } + {[10]} {'Exx' } {'Gr_I '} {'Gr_I '} {'0.0089104' } {'0.0053424' } + {[22]} {'Exx' } {'Gr_I '} {'Infl ' } {'0.00063139' } {'0.00064897' } + {[23]} {'Exx' } {'Gr_I '} {'r1 ' } {'0.0011031' } {'0.0019533' } + {[24]} {'Exx' } {'Gr_I '} {'r40 ' } {'0.0018445' } {'0.0020602' } + {[25]} {'Exx' } {'Gr_I '} {'xhr40 '} {'0.00095556' } {'0.0064856' } + {[11]} {'Exx' } {'Infl ' } {'Infl ' } {'0.0020268' } {'0.0020922' } + {[26]} {'Exx' } {'Infl ' } {'r1 ' } {'0.0025263' } {'0.0036375' } + {[27]} {'Exx' } {'Infl ' } {'r40 ' } {'0.0029126' } {'0.0034139' } + {[28]} {'Exx' } {'Infl ' } {'xhr40 '} {'-0.00077101'} {'-0.0011665' } + {[12]} {'Exx' } {'r1 ' } {'r1 ' } {'0.0038708' } {'0.0066074' } + {[29]} {'Exx' } {'r1 ' } {'r40 ' } {'0.0044773' } {'0.0062959' } + {[30]} {'Exx' } {'r1 ' } {'xhr40 '} {'-0.00048202'} {'-0.00075499'} + {[13]} {'Exx' } {'r40 ' } {'r40 ' } {'0.0054664' } {'0.0061801' } + {[31]} {'Exx' } {'r40 ' } {'xhr40 '} {'0.00053864' } {'-0.00030456'} + {[14]} {'Exx' } {'xhr40 '} {'xhr40 '} {'0.053097' } {'0.012048' } + {[15]} {'Exx' } {'GoY '} {'GoY '} {'2.4863' } {'2.4872' } + {[16]} {'Exx' } {'hours '} {'hours '} {'0.0018799' } {'0.0018759' } + {[32]} {'Exx1'} {'Gr_C '} {'Gr_C '} {'0.00077917' } {'0.00080528' } + {[33]} {'Exx1'} {'Gr_I '} {'Gr_I '} {'0.0050104' } {'0.0017036' } + {[34]} {'Exx1'} {'Infl ' } {'Infl ' } {'0.0019503' } {'0.0020185' } + {[35]} {'Exx1'} {'r1 ' } {'r1 ' } {'0.0038509' } {'0.0065788' } + {[36]} {'Exx1'} {'r40 ' } {'r40 ' } {'0.0054699' } {'0.0061762' } + {[37]} {'Exx1'} {'xhr40 '} {'xhr40 '} {'-0.00098295'} {'-4.5519e-15'} + {[38]} {'Exx1'} {'GoY '} {'GoY '} {'2.4868' } {'2.4863' } + {[39]} {'Exx1'} {'hours '} {'hours '} {'0.0018799' } {'0.0018755' } +]; + +% Output of the replication files for orderApp=2 +AndreasenEtAl.Q2 = 140.8954; +AndreasenEtAl.moments2 =[ % note that we reshuffeled to be compatible with our matched moments block + {[ 1]} {'Ex' } {'Gr_C '} {' ' } {'0.024388' } {'0.023726' } + {[ 2]} {'Ex' } {'Gr_I '} {' ' } {'0.031046' } {'0.027372' } + {[ 3]} {'Ex' } {'Infl ' } {' ' } {'0.03757' } {'0.034618' } + {[ 4]} {'Ex' } {'r1 ' } {' ' } {'0.056048' } {'0.056437' } + {[ 5]} {'Ex' } {'r40 ' } {' ' } {'0.069929' } {'0.07051' } + {[ 6]} {'Ex' } {'xhr40 '} {' ' } {'0.017237' } {'0.014242' } + {[ 7]} {'Ex' } {'GoY '} {' ' } {'-1.5745' } {'-1.574' } + {[ 8]} {'Ex' } {'hours '} {' ' } {'-0.043353' } {'-0.043351' } + {[ 9]} {'Exx' } {'Gr_C '} {'Gr_C '} {'0.0013159' } {'0.0012917' } + {[17]} {'Exx' } {'Gr_C '} {'Gr_I '} {'0.0021789' } {'0.0017862' } + {[18]} {'Exx' } {'Gr_C '} {'Infl ' } {'0.00067495' } {'0.00061078' } + {[19]} {'Exx' } {'Gr_C '} {'r1 ' } {'0.0011655' } {'0.0011494' } + {[20]} {'Exx' } {'Gr_C '} {'r40 ' } {'0.0015906' } {'0.0016149' } + {[21]} {'Exx' } {'Gr_C '} {'xhr40 '} {'0.0020911' } {'0.002203' } + {[10]} {'Exx' } {'Gr_I '} {'Gr_I '} {'0.0089104' } {'0.0054317' } + {[22]} {'Exx' } {'Gr_I '} {'Infl ' } {'0.00063139' } {'0.00045278' } + {[23]} {'Exx' } {'Gr_I '} {'r1 ' } {'0.0011031' } {'0.0013672' } + {[24]} {'Exx' } {'Gr_I '} {'r40 ' } {'0.0018445' } {'0.0018557' } + {[25]} {'Exx' } {'Gr_I '} {'xhr40 '} {'0.00095556' } {'0.0067742' } + {[11]} {'Exx' } {'Infl ' } {'Infl ' } {'0.0020268' } {'0.0016583' } + {[26]} {'Exx' } {'Infl ' } {'r1 ' } {'0.0025263' } {'0.0024521' } + {[27]} {'Exx' } {'Infl ' } {'r40 ' } {'0.0029126' } {'0.002705' } + {[28]} {'Exx' } {'Infl ' } {'xhr40 '} {'-0.00077101'} {'-0.00065007'} + {[12]} {'Exx' } {'r1 ' } {'r1 ' } {'0.0038708' } {'0.0038274' } + {[29]} {'Exx' } {'r1 ' } {'r40 ' } {'0.0044773' } {'0.004297' } + {[30]} {'Exx' } {'r1 ' } {'xhr40 '} {'-0.00048202'} {'6.3243e-05' } + {[13]} {'Exx' } {'r40 ' } {'r40 ' } {'0.0054664' } {'0.0051686' } + {[31]} {'Exx' } {'r40 ' } {'xhr40 '} {'0.00053864' } {'0.00066645' } + {[14]} {'Exx' } {'xhr40 '} {'xhr40 '} {'0.053097' } {'0.013543' } + {[15]} {'Exx' } {'GoY '} {'GoY '} {'2.4863' } {'2.4858' } + {[16]} {'Exx' } {'hours '} {'hours '} {'0.0018799' } {'0.0018804' } + {[32]} {'Exx1'} {'Gr_C '} {'Gr_C '} {'0.00077917' } {'0.00081772' } + {[33]} {'Exx1'} {'Gr_I '} {'Gr_I '} {'0.0050104' } {'0.0017106' } + {[34]} {'Exx1'} {'Infl ' } {'Infl ' } {'0.0019503' } {'0.0015835' } + {[35]} {'Exx1'} {'r1 ' } {'r1 ' } {'0.0038509' } {'0.0037985' } + {[36]} {'Exx1'} {'r40 ' } {'r40 ' } {'0.0054699' } {'0.0051642' } + {[37]} {'Exx1'} {'xhr40 '} {'xhr40 '} {'-0.00098295'} {'0.00020285' } + {[38]} {'Exx1'} {'GoY '} {'GoY '} {'2.4868' } {'2.4848' } + {[39]} {'Exx1'} {'hours '} {'hours '} {'0.0018799' } {'0.0018799' } +]; + +@#for orderApp in 1:2 + +method_of_moments( + mom_method = GMM % method of moments method; possible values: GMM|SMM + , datafile = 'AFVRR_data.mat' % name of filename with data + , bartlett_kernel_lag = 10 % bandwith in optimal weighting matrix + , order = @{orderApp} % order of Taylor approximation in perturbation + , pruning % use pruned state space system at higher-order + % , verbose % display and store intermediate estimation results + , weighting_matrix = ['DIAGONAL'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename + % , TeX % print TeX tables and graphics + % Optimization options that can be set by the user in the mod file, otherwise default values are provided + %, huge_number=1D10 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons + , mode_compute = 0 % specifies the optimizer for minimization of moments distance, note that by default there is a new optimizer + , optim = ('TolFun', 1e-6 + ,'TolX', 1e-6 + ,'MaxIter', 3000 + ,'MaxFunEvals', 1D6 + ,'UseParallel' , 1 + %,'Jacobian' , 'on' + ) % a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute + %, silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + %, analytic_standard_errors + , se_tolx=1e-10 +); + +% Check results + +fprintf('****************************************************************\n') +fprintf('Compare Results for perturbation order @{orderApp}\n') +fprintf('****************************************************************\n') +dev_Q = AndreasenEtAl.Q@{orderApp} - oo_.mom.Q; +dev_datamoments = str2double(AndreasenEtAl.moments@{orderApp}(:,5)) - oo_.mom.data_moments; +dev_modelmoments = str2double(AndreasenEtAl.moments@{orderApp}(:,6)) - oo_.mom.model_moments; + +table([AndreasenEtAl.Q@{orderApp} ; str2double(AndreasenEtAl.moments@{orderApp}(:,5)) ; str2double(AndreasenEtAl.moments@{orderApp}(:,6))],... + [oo_.mom.Q ; oo_.mom.data_moments ; oo_.mom.model_moments ],... + [dev_Q ; dev_datamoments ; dev_modelmoments ],... + 'VariableNames', {'Andreasen et al', 'Dynare', 'dev'},... + 'RowNames', ['Q'; strcat('Data_', M_.matched_moments(:,4)); strcat('Model_', M_.matched_moments(:,4))]) + +if norm(dev_modelmoments)> 1e-4 + warning('Something wrong in the computation of moments at order @{orderApp}') +end + +@#endfor + +%-------------------------------------------------------------------------- +% Replicate estimation at orderApp=3 +%-------------------------------------------------------------------------- +@#ifdef DoEstimation +method_of_moments( + mom_method = GMM % method of moments method; possible values: GMM|SMM + , datafile = 'AFVRR_data.mat' % name of filename with data + , bartlett_kernel_lag = 10 % bandwith in optimal weighting matrix + , order = 3 % order of Taylor approximation in perturbation + , pruning % use pruned state space system at higher-order + % , verbose % display and store intermediate estimation results + , weighting_matrix = ['DIAGONAL', 'Optimal'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename + % , TeX % print TeX tables and graphics + % Optimization options that can be set by the user in the mod file, otherwise default values are provided + %, huge_number=1D10 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons + , mode_compute = 13 % specifies the optimizer for minimization of moments distance, note that by default there is a new optimizer + , additional_optimizer_steps = [13] + , optim = ('TolFun', 1e-6 + ,'TolX', 1e-6 + ,'MaxIter', 3000 + ,'MaxFunEvals', 1D6 + ,'UseParallel' , 1 + %,'Jacobian' , 'on' + ) % a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute + %, silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + %, analytic_standard_errors + , se_tolx=1e-10 +); +@#endif \ No newline at end of file diff --git a/tests/estimation/method_of_moments/AFVRR/AFVRR_common.inc b/tests/estimation/method_of_moments/AFVRR/AFVRR_common.inc new file mode 100644 index 000000000..76aea9e0b --- /dev/null +++ b/tests/estimation/method_of_moments/AFVRR/AFVRR_common.inc @@ -0,0 +1,540 @@ +% DSGE model based on replication files of +% Andreasen, Fernandez-Villaverde, Rubio-Ramirez (2018), The Pruned State-Space System for Non-Linear DSGE Models: Theory and Empirical Applications, Review of Economic Studies, 85, p. 1-49 +% Original code by Martin M. Andreasen, Jan 2016 +% Adapted for Dynare by Willi Mutschler (@wmutschl, willi@mutschler.eu), Jan 2021 +% ========================================================================= +% Copyright (C) 2021 Dynare Team +% +% This file is part of Dynare. +% +% Dynare is free software: you can redistribute it and/or modify +% it under the terms of the GNU General Public License as published by +% the Free Software Foundation, either version 3 of the License, or +% (at your option) any later version. +% +% Dynare is distributed in the hope that it will be useful, +% but WITHOUT ANY WARRANTY; without even the implied warranty of +% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +% GNU General Public License for more details. +% +% You should have received a copy of the GNU General Public License +% along with Dynare. If not, see . +% ========================================================================= + +%-------------------------------------------------------------------------- +% Variable declaration +%-------------------------------------------------------------------------- +var +ln_k +ln_s +ln_a +ln_g +ln_d + +ln_c +ln_r +ln_pai +ln_h +ln_q +ln_evf +ln_iv +ln_x2 +ln_la +ln_goy +ln_Esdf + +xhr20 +xhr40 +Exhr + +@#for i in 1:40 +ln_p@{i} +@#endfor + +Obs_Gr_C +Obs_Gr_I +Obs_Infl +Obs_r1 +Obs_r40 +Obs_xhr40 +Obs_GoY +Obs_hours +; + +predetermined_variables ln_k ln_s; + +varobs Obs_Gr_C Obs_Gr_I Obs_Infl Obs_r1 Obs_r40 Obs_xhr40 Obs_GoY Obs_hours; + +%-------------------------------------------------------------------------- +% Exogenous shocks +%-------------------------------------------------------------------------- +varexo +eps_a +eps_d +eps_g +; + +%-------------------------------------------------------------------------- +% Parameter declaration +%-------------------------------------------------------------------------- +parameters +BETTA +B +INHABIT +H +PHI1 +PHI2 +RRA +PHI4 +KAPAone +KAPAtwo +DELTA +THETA +ETA +ALFA +CHI +RHOR +BETTAPAI +BETTAY +MYYPS +MYZ +RHOA +%STDA +RHOG +%STDG +RHOD +%STDD +CONSxhr40 +BETTAxhr +BETTAxhr40 +CONSxhr20 +PAI +GAMA +GoY + +%auxiliary +PHIzero +AA +PHI3 +negVf +; + + +%-------------------------------------------------------------------------- +% Model equations +%-------------------------------------------------------------------------- +% Based on DSGE_model_NegVf_yieldCurve.m and DSGE_model_PosVf_yieldCurve.m +% Note that we include an auxiliary parameter negVf to distinguish whether +% the steady state value function is positive (negVf=0) or negative (negVf=1). +% This parameter is endogenously determined in the steady_state_model block. + +model; +%-------------------------------------------------------------------------- +% Auxiliary expressions +%-------------------------------------------------------------------------- +% do exp transform such that variables are logged variables +@#for var in [ "k", "s", "c", "r", "a", "g", "d", "pai", "h", "q", "evf", "iv", "x2", "la", "goy", "Esdf" ] +#@{var}_ba1 = exp(ln_@{var}(-1)); +#@{var}_cu = exp(ln_@{var}); +#@{var}_cup = exp(ln_@{var}(+1)); +@#endfor +@#for i in 1:40 +#p@{i}_cu = exp(ln_p@{i}); +#p@{i}_cup = exp(ln_p@{i}(+1)); +@#endfor +% these variables are not transformed +#xhr20_cu = xhr20; +#xhr20_cup = xhr20(+1); +#xhr40_cu = xhr40; +#xhr40_cup = xhr40(+1); +#Exhr_cu = Exhr; +#Exhr_cup = Exhr(+1); + +% auxiliary steady state variables +#K = exp(steady_state(ln_k)); +#IV = exp(steady_state(ln_iv)); +#C = exp(steady_state(ln_c)); +#Y = (C + IV)/(1-GoY); +#R = exp(steady_state(ln_r)); +#G = Y-C-IV; + +#removeMeanXhr = 1; + +% The atemporal relations if possible +% No stochastic trend in investment specific shocks +#myyps_cu = MYYPS; +#myyps_cup = MYYPS; + +% No stochastic trend in non-stationary technology shocks +#myz_cu = MYZ; +#myz_cup = MYZ; + +% Defining myzstar +#MYZSTAR = MYYPS^(THETA/(1-THETA))*MYZ; +#myzstar_cu = myyps_cu ^(THETA/(1-THETA))*myz_cu; +#myzstar_cup= myyps_cup^(THETA/(1-THETA))*myz_cup; + +% The expression for the value function (only valid for deterministic trends!) +% Note that we make use of auxiliary parameter negVf to switch signs +#mvf_cup = -negVf*(d_cup/(1-PHI2)*((c_cup-B*c_cu*MYZSTAR^-1)^(1-PHI2)-1) + d_cup*PHIzero/(1-PHI1)*(1-h_cup)^(1-PHI1) - negVf* BETTA*MYZSTAR^((1-PHI4)*(1-PHI2))*AA*evf_cup^(1/(1-PHI3))); + +% The growth rate in lambda +#myla_cup = (la_cup/la_cu)*(AA*evf_cu^(1/(1-PHI3))/mvf_cup)^PHI3*myzstar_cup^(-PHI2*(1-PHI4)-PHI4); + +% The relation between the optimal price for the firms and the pris and inflation +%ptil_cu = ((1-ALFA*(pai_ba1^CHI/pai_cu )^(1-ETA))/(1-ALFA))^(1/(1-ETA)); +%ptil_cup = ((1-ALFA*(pai_cu ^CHI/pai_cup)^(1-ETA))/(1-ALFA))^(1/(1-ETA)); +#ptil_cu = ((1-ALFA*(1/pai_cu )^(1-ETA))/(1-ALFA))^(1/(1-ETA)); +#ptil_cup = ((1-ALFA*(1/pai_cup)^(1-ETA))/(1-ALFA))^(1/(1-ETA)); + +% From the households' FOC for labor +#w_cu = d_cu*PHIzero*(1-h_cu )^(-PHI1)/la_cu; +#w_cup = d_cu*PHIzero*(1-h_cup)^(-PHI1)/la_cup; +% Shouldn't w_cup include d_cup? Let's stick to the original (wrong) code in the replication files as results don't change dramatically... [@wmutschl] + +% The firms' FOC for labor +#mc_cu = w_cu /((1-THETA)*a_cu *myyps_cu ^(-THETA/(1-THETA))*myz_cu ^-THETA *k_cu ^THETA*h_cu ^(-THETA)); +#mc_cup = w_cup/((1-THETA)*a_cup*myyps_cup^(-THETA/(1-THETA))*myz_cup^-THETA *k_cup^THETA*h_cup^(-THETA)); + +% The firms' FOC for capital +#rk_cu = mc_cu *THETA* a_cu *myyps_cu *myz_cu ^(1-THETA)*k_cu ^(THETA-1)*h_cu ^(1-THETA); +#rk_cup = mc_cup*THETA* a_cup*myyps_cup*myz_cup^(1-THETA)*k_cup^(THETA-1)*h_cup^(1-THETA); + +% The income identity +#y_cu = c_cu + iv_cu + g_cu; + +%-------------------------------------------------------------------------- +% Actual model equations +%-------------------------------------------------------------------------- + +[name='Expected value of the value function'] +0 = -evf_cu + (mvf_cup/AA)^(1-PHI3); + +[name='Households FOC for capital'] +0 = -q_cu+BETTA*myla_cup/myyps_cup*(rk_cup+q_cup*(1-DELTA) -q_cup*KAPAtwo/2*(iv_cup/k_cup*myyps_cup*myzstar_cup - IV/K*MYYPS*MYZSTAR)^2 +q_cup*KAPAtwo*(iv_cup/k_cup*myyps_cup*myzstar_cup - IV/K*MYYPS*MYZSTAR)*iv_cup/k_cup*myyps_cup*myzstar_cup); + +[name='Households FOC for investments'] +0 = -1+q_cu*(1-KAPAone/2*(iv_cu/IV-1)^2-iv_cu/IV*KAPAone*(iv_cu/IV-1)-KAPAtwo*(iv_cu/k_cu*myyps_cu*myzstar_cu - IV/K*MYYPS*MYZSTAR)); + +[name='Euler equation for consumption'] +0 = -1+BETTA*r_cu*exp(CONSxhr40*xhr40_cu + CONSxhr20*xhr20_cu)*myla_cup/pai_cup; + +[name='Households FOC for consumption'] +0 = -la_cu + d_cu*(c_cu -B*c_ba1*myzstar_cu^-1)^(-PHI2) -INHABIT*B*BETTA*d_cup*(AA*evf_cu^(1/(1-PHI3))/mvf_cup)^PHI3*(c_cup -B*c_cu*myzstar_cup^-1)^(-PHI2)*myzstar_cup^(-PHI2*(1-PHI4)-PHI4); + +[name='Nonlinear pricing, relation for x1 = (ETA-1)/ETA*x2'] +0= -(ETA-1)/ETA*x2_cu+y_cu*mc_cu*ptil_cu^(-ETA-1) +ALFA*BETTA*myla_cup*(ptil_cu/ptil_cup)^(-ETA-1)*(1/pai_cup)^(-ETA)*(ETA-1)/ETA*x2_cup*myzstar_cup; + +[name='Nonlinear pricing, relation for x2'] +0=-x2_cu+y_cu*ptil_cu^-ETA +ALFA*BETTA*myla_cup*(ptil_cu/ptil_cup)^(-ETA)*(1/pai_cup)^(1-ETA)*x2_cup*myzstar_cup; + +[name='Nonlinear pricing, relation for s'] +0= -s_cup+(1-ALFA)*ptil_cu^(-ETA)+ALFA*(pai_cu/1)^ETA*s_cu; + +[name='Interest rate rule'] +0 = -log(r_cu/R)+RHOR*log(r_ba1/R)+(1-RHOR)*(BETTAPAI*log(pai_cu/PAI)+BETTAY*log(y_cu/Y) + BETTAxhr*(BETTAxhr40*xhr40_cu - removeMeanXhr*Exhr_cu)); + +[name='Production function'] +0 = -y_cu*s_cup + a_cu *(k_cu *myyps_cu ^(-1/(1-THETA))*myz_cu ^-1)^THETA*h_cu ^(1-THETA); + +[name='Relation for physical capital stock'] +0= -k_cup + (1-DELTA)*k_cu*(myyps_cu*myzstar_cu)^-1 + iv_cu - iv_cu*KAPAone/2*(iv_cu/IV-1)^2 - k_cu*(myyps_cu*myzstar_cu)^-1*KAPAtwo/2*(iv_cu/k_cu*myyps_cu*myzstar_cu - IV/K*MYYPS*MYZSTAR)^2; + +[name='Goverment spending over output'] +0=-goy_cu + g_cu/y_cu; + +[name='The yield curve: p1'] +0= -p1_cu + 1/r_cu; + +@#for i in 2:40 +[name='The yield curve: p@{i}'] +0= -p@{i}_cu + BETTA*myla_cup/pai_cup*p@{i-1}_cup; +@#endfor + +[name='Stochastic discount factor'] +0= -Esdf_cu+ BETTA*myla_cup/pai_cup; + +[name='Expected 5 year excess holding period return'] +0= -xhr20_cu+ log(p19_cup) - log(p20_cu) - log(r_cu); + +[name='Expected 10 year excess holding period return'] +0= -xhr40_cu+ log(p39_cup) - log(p40_cu) - log(r_cu); + +[name='Mean of expected excess holding period return in Taylor rule'] +0= -Exhr_cu + (1-GAMA)*(BETTAxhr40*xhr40_cu) + GAMA*Exhr_cup; + +[name='Exogenous process for productivity'] +0 = -log(a_cu)+RHOA*log(a_ba1) + eps_a; + +[name='Exogenous process for government spending'] +0 = -log(g_cu/G)+RHOG*log(g_ba1/G) + eps_g; + +[name='Exogenous process for discount factor shifter'] +0 = -log(d_cu)+RHOD*log(d_ba1) + eps_d; + +[name='Observable annualized consumption growth'] +Obs_Gr_C = 4*( ln_c -ln_c(-1) + log(MYZSTAR)); + +[name='Observable annualized investment growth'] +Obs_Gr_I = 4*( ln_iv - ln_iv(-1) + log(MYZSTAR)+log(MYYPS)); + +[name='Observable annualized inflation'] +Obs_Infl = 4*( ln_pai); + +[name='Observable annualized one-quarter nominal yield'] +Obs_r1 = 4*( ln_r); + +[name='Observable annualized 10-year nominal yield'] +Obs_r40 = 4*( -1/40*ln_p40); + +[name='Observable annualized 10-year ex post excess holding period return'] +Obs_xhr40 = 4*( ln_p39 - ln_p40(-1) - ln_r(-1) ); + +[name='Observable annualized log ratio of government spending to GDP'] +Obs_GoY = 4*( 1/4*ln_goy); + +[name='Observable annualized log of hours'] +Obs_hours = 4*( 1/100*ln_h); +end; + + +%-------------------------------------------------------------------------- +% Steady State Computations +%-------------------------------------------------------------------------- +% Based on DSGE_model_yieldCurve_ss.m, getPHI3.m, ObjectGMM.m +% Note that we include an auxiliary parameter negVf to distinguish whether +% the steady state value function is positive (negVf=0) or negative (negVf=1). +% This parameter is endogenously determined in the steady_state_model block. + + +steady_state_model; + +% The growth rate in the firms' fixed costs +MYZSTARBAR = MYYPS^(THETA/(1-THETA))*MYZ; + +% The growth rate for lampda +MYLABAR = MYZSTARBAR^(-PHI2*(1-PHI4)-PHI4); + +% The relative optimal price for firms +PTILBAR = ((1-ALFA*PAI^((CHI-1)*(1-ETA)))/(1-ALFA))^(1/(1-ETA)); + +% The state variable s for distortions between output and produktion +SBAR = ((1-ALFA)*PTILBAR^(-ETA))/(1-ALFA*PAI^((1-CHI)*ETA)); + +% The 1-period interest rate +RBAR = PAI/(BETTA*MYLABAR); + +% The market price of capital +QBAR = 1; + +% The real price of renting capital +RKBAR = QBAR*(MYYPS/(BETTA*MYLABAR)-(1-DELTA)); + +% The marginal costs in the firms +MCBAR = (1-ALFA*BETTA*MYLABAR*PAI^((1-CHI)*ETA)*MYZSTARBAR)*(ETA-1)/ETA*PTILBAR/(1-ALFA*BETTA*MYLABAR*PAI^((CHI-1)*(1-ETA))*MYZSTARBAR); + +% The capital stock +KBAR = H*(RKBAR/(MCBAR*THETA*MYYPS*MYZ^(1-THETA)))^(1/(THETA-1)); + +% The wage level +WBAR = MCBAR*(1-THETA)*MYYPS^(-THETA/(1-THETA))*MYZ^-THETA*(KBAR/H)^THETA; + +% The level of investment +IVBAR = KBAR - (1-DELTA)*KBAR*MYYPS^(-1/(1-THETA))*MYZ^-1; + +% The consumption level +CBAR = ((1-GoY)*(KBAR*MYYPS^(-1/(1-THETA))*MYZ^-1)^THETA*H^(1-THETA))/SBAR-IVBAR; + +% The output level +YBAR = (CBAR + IVBAR)/(1-GoY); + +% The value of lambda +LABAR = (CBAR-B*CBAR*MYZSTARBAR^-1)^-PHI2 - INHABIT*B*BETTA*(CBAR-B*CBAR*MYZSTARBAR^-1)^-PHI2*MYZSTARBAR^(-PHI2*(1-PHI4)-PHI4); + +% The value of PHIzero +PHIzero = LABAR*WBAR*(1-H)^PHI1; + +% The level of the value function +VFBAR = 1/(1-BETTA*MYZSTARBAR^((1-PHI4)*(1-PHI2)))*(1/(1-PHI2)*((CBAR-B*CBAR*MYZSTARBAR^-1)^(1-PHI2)-1)+PHIzero/(1-PHI1)*(1-H)^(1-PHI1)); +UBAR = 1/(1-PHI2)*((CBAR-B*CBAR*MYZSTARBAR^-1)^(1-PHI2)-1)+PHIzero/(1-PHI1)*(1-H)^(1-PHI1); +[AA, EVFBAR, PHI3, negVf, info]= AFVRR_steady_helper(VFBAR,RBAR,IVBAR,CBAR,KBAR,LABAR,QBAR,YBAR, BETTA,B,PAI,H,PHIzero,PHI1,PHI2,THETA,MYYPS,MYZ,INHABIT,RRA,CONSxhr40); +% The value of X2 +X2BAR = YBAR*PTILBAR^(-ETA)/(1-BETTA*ALFA*MYLABAR*PAI^((CHI-1)*(1-ETA))*MYZSTARBAR); + +% Government spending +GBAR = GoY*YBAR; +%************************************************************************** + +% map into model variables +ln_k = log(KBAR); +ln_s = log(SBAR); +ln_c_ba1 = log(CBAR); +ln_r_ba1 = log(RBAR); +ln_a = log(1); +ln_g = log(GBAR); +ln_d = log(1); + +ln_c = log(CBAR); +ln_r = log(RBAR); +ln_pai = log(PAI); +ln_h = log(H); +ln_q = log(QBAR); +ln_evf = log(EVFBAR); +ln_iv = log(IVBAR); +ln_x2 = log(X2BAR); +ln_la = log(LABAR); +ln_goy = log(GoY); +ln_Esdf = log(1/RBAR); +xhr20 = 0; +xhr40 = 0; +Exhr = 0; +% The yield curve +ln_p1 = log((1/RBAR)^1); +ln_p2 = log((1/RBAR)^2); +ln_p3 = log((1/RBAR)^3); +ln_p4 = log((1/RBAR)^4); +ln_p5 = log((1/RBAR)^5); +ln_p6 = log((1/RBAR)^6); +ln_p7 = log((1/RBAR)^7); +ln_p8 = log((1/RBAR)^8); +ln_p9 = log((1/RBAR)^9); +ln_p10 = log((1/RBAR)^10); +ln_p11 = log((1/RBAR)^11); +ln_p12 = log((1/RBAR)^12); +ln_p13 = log((1/RBAR)^13); +ln_p14 = log((1/RBAR)^14); +ln_p15 = log((1/RBAR)^15); +ln_p16 = log((1/RBAR)^16); +ln_p17 = log((1/RBAR)^17); +ln_p18 = log((1/RBAR)^18); +ln_p19 = log((1/RBAR)^19); +ln_p20 = log((1/RBAR)^20); +ln_p21 = log((1/RBAR)^21); +ln_p22 = log((1/RBAR)^22); +ln_p23 = log((1/RBAR)^23); +ln_p24 = log((1/RBAR)^24); +ln_p25 = log((1/RBAR)^25); +ln_p26 = log((1/RBAR)^26); +ln_p27 = log((1/RBAR)^27); +ln_p28 = log((1/RBAR)^28); +ln_p29 = log((1/RBAR)^29); +ln_p30 = log((1/RBAR)^30); +ln_p31 = log((1/RBAR)^31); +ln_p32 = log((1/RBAR)^32); +ln_p33 = log((1/RBAR)^33); +ln_p34 = log((1/RBAR)^34); +ln_p35 = log((1/RBAR)^35); +ln_p36 = log((1/RBAR)^36); +ln_p37 = log((1/RBAR)^37); +ln_p38 = log((1/RBAR)^38); +ln_p39 = log((1/RBAR)^39); +ln_p40 = log((1/RBAR)^40); + +Obs_Gr_C = 4*( log(MYZSTARBAR) ); +Obs_Gr_I = 4*( log(MYZSTARBAR)+log(MYYPS) ); +Obs_Infl = 4*( ln_pai ); +Obs_r1 = 4*( ln_r ); +Obs_r40 = 4*( -1/40*ln_p40 ); +Obs_xhr40 = 4*( xhr40 ); +Obs_GoY = 4*( 1/4*ln_goy ); +Obs_hours = 4*( 1/100*ln_h ); +end; + +%-------------------------------------------------------------------------- +% Declare moments to use in estimation +%-------------------------------------------------------------------------- +% These are the moments used in the paper; corresponds to momentSet=2 in the replication files + +matched_moments; +%mean +Obs_Gr_C; +Obs_Gr_I; +Obs_Infl; +Obs_r1; +Obs_r40; +Obs_xhr40; +Obs_GoY; +Obs_hours; + +% all variances +Obs_Gr_C*Obs_Gr_C; +Obs_Gr_I*Obs_Gr_I; +Obs_Infl*Obs_Infl; +Obs_r1*Obs_r1; +Obs_r40*Obs_r40; +Obs_xhr40*Obs_xhr40; +Obs_GoY*Obs_GoY; +Obs_hours*Obs_hours; + +% covariance excluding GoY and hours +Obs_Gr_C*Obs_Gr_I; +Obs_Gr_C*Obs_Infl; +Obs_Gr_C*Obs_r1; +Obs_Gr_C*Obs_r40; +Obs_Gr_C*Obs_xhr40; +%Obs_Gr_C*Obs_GoY; +%Obs_Gr_C*Obs_hours; + +Obs_Gr_I*Obs_Infl; +Obs_Gr_I*Obs_r1; +Obs_Gr_I*Obs_r40; +Obs_Gr_I*Obs_xhr40; +%Obs_Gr_I*Obs_GoY; +%Obs_Gr_I*Obs_hours; + +Obs_Infl*Obs_r1; +Obs_Infl*Obs_r40; +Obs_Infl*Obs_xhr40; +%Obs_Infl*Obs_GoY; +%Obs_Infl*Obs_hours; + +Obs_r1*Obs_r40; +Obs_r1*Obs_xhr40; +%Obs_r1*Obs_GoY; +%Obs_r1*Obs_hours; + +Obs_r40*Obs_xhr40; +%Obs_r40*Obs_GoY; +%Obs_r40*Obs_hours; + +%Obs_xhr40*Obs_GoY; +%Obs_xhr40*Obs_hours; + +%Obs_GoY*Obs_hours; + +%first autocovariance +Obs_Gr_C*Obs_Gr_C(-1); +Obs_Gr_I*Obs_Gr_I(-1); +Obs_Infl*Obs_Infl(-1); +Obs_r1*Obs_r1(-1); +Obs_r40*Obs_r40(-1); +Obs_xhr40*Obs_xhr40(-1); +Obs_GoY*Obs_GoY(-1); +Obs_hours*Obs_hours(-1); +end; + +%-------------------------------------------------------------------------- +% Create Data +%-------------------------------------------------------------------------- +@#ifdef CreateData +verbatim; +% From 1961Q3 to 2007Q4 +DataUS = xlsread('Data_PruningPaper_v5.xlsx','Data_used','E3:M188'); +% ANNUALIZED (except for hours and GoY) +% 1 2 3 4 5 6 7 8 9 +% Lables: Date Gr_C Gr_I GoY hours Infl_C r1 r40 xhr40 +%label_data = {'Gr_C ', 'Gr_I ','Infl ', 'r1 ', 'r40 ', 'xhr40 ','GoY ', 'hours '}; +%DataUS = [DataUS(:,2:3) DataUS(:,6:8) DataUS(:,9) log(DataUS(:,4)) 4*log(DataUS(:,5))/100]; +Obs_Gr_C = DataUS(:,2); +Obs_Gr_I = DataUS(:,3); +Obs_Infl = DataUS(:,6); +Obs_r1 = DataUS(:,7); +Obs_r40 = DataUS(:,8); +Obs_xhr40 = DataUS(:,9); +Obs_GoY = log(DataUS(:,4)); +Obs_hours = 4*log(DataUS(:,5))/100; + +save('AFVRR_data.mat','Obs_Gr_C','Obs_Gr_I','Obs_Infl','Obs_r1','Obs_r40','Obs_xhr40','Obs_GoY','Obs_hours'); +pause(1); +end; +@#endif \ No newline at end of file diff --git a/tests/estimation/method_of_moments/AFVRR/AFVRR_data.mat b/tests/estimation/method_of_moments/AFVRR/AFVRR_data.mat new file mode 100644 index 000000000..f606b2109 Binary files /dev/null and b/tests/estimation/method_of_moments/AFVRR/AFVRR_data.mat differ diff --git a/tests/estimation/method_of_moments/AFVRR/AFVRR_steady_helper.m b/tests/estimation/method_of_moments/AFVRR/AFVRR_steady_helper.m new file mode 100644 index 000000000..b8289d484 --- /dev/null +++ b/tests/estimation/method_of_moments/AFVRR/AFVRR_steady_helper.m @@ -0,0 +1,80 @@ +% DSGE model based on replication files of +% Andreasen, Fernandez-Villaverde, Rubio-Ramirez (2018), The Pruned State-Space System for Non-Linear DSGE Models: Theory and Empirical Applications, Review of Economic Studies, 85, p. 1-49 +% Adapted for Dynare by Willi Mutschler (@wmutschl, willi@mutschler.eu), Jan 2021 +% ========================================================================= +% Copyright (C) 2021 Dynare Team +% +% This file is part of Dynare. +% +% Dynare is free software: you can redistribute it and/or modify +% it under the terms of the GNU General Public License as published by +% the Free Software Foundation, either version 3 of the License, or +% (at your option) any later version. +% +% Dynare is distributed in the hope that it will be useful, +% but WITHOUT ANY WARRANTY; without even the implied warranty of +% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +% GNU General Public License for more details. +% +% You should have received a copy of the GNU General Public License +% along with Dynare. If not, see . +% ========================================================================= + +% This is a helper function to compute steady state values and endogenous parameters +% Based on DSGE_model_yieldCurve_ss.m, getPHI3.m, ObjectGMM.m + +function [AA, EVFBAR, PHI3, negVf, info]= AFVRR_steady_helper(VFBAR,RBAR,IVBAR,CBAR,KBAR,LABAR,QBAR,YBAR, BETTA,B,PAI,H,PHIzero,PHI1,PHI2,THETA,MYYPS,MYZ,INHABIT,RRA,CONSxhr40) +% We get nice values of EVF by setting AA app. equal to VF. +% The value of the expected value function raised to the power 1-PHI3 +% Also we check bounds on other variables +% % Adding PHI3 to params. Note that PHI3 only affects the value function in +% % steady state, hence the value we assign to PHI3 is irrelevant +% PHI3 = -100; + +info=0; +AA = NaN; +EVFBAR = NaN; +PHI3 = NaN; +negVf = NaN; + +MYZSTAR = MYYPS^(THETA/(1-THETA))*MYZ; +% The wage level +WBAR = PHIzero*(1-H)^(-PHI1)/LABAR; +RRAc = RRA; +if INHABIT == 1 + PHI3 = (RRAc - PHI2/((1-B*MYZSTAR^-1)/(1-BETTA*B)+PHI2/PHI1*WBAR*(1-H)/CBAR))/((1-PHI2)/((1-B*MYZSTAR^-1)/(1-BETTA*B)-(CBAR-B*CBAR*MYZSTAR^-1)^PHI2/((1-BETTA*B)*CBAR)+WBAR*(1-H)/CBAR*(1-PHI2)/(1-PHI1))); +else + PHI3 = (RRAc - PHI2/(1-B*MYZSTAR^-1+PHI2/PHI1*WBAR*(1-H)/CBAR))/((1-PHI2)/(1-B*MYZSTAR^-1-(CBAR-B*CBAR*MYZSTAR^-1)^PHI2/((1-BETTA*B)*CBAR)+WBAR*(1-H)/CBAR*(1-PHI2)/(1-PHI1))); +end +if abs(PHI3) > 30000 + disp('abs of PHI3 exceeds 30000') + info=1; + return +end + +if CONSxhr40 > 1 + info=1; + return +end + + +if VFBAR < 0 + AA = -VFBAR; + EVFBAR = (-VFBAR/AA)^(1-PHI3); + negVf = 1; +else + AA = VFBAR; + EVFBAR = (VFBAR/AA)^(1-PHI3); + negVf = -1; + disp('Positive Value Function'); +end + + +if RBAR < 1 || IVBAR < 0 || CBAR < 0 || KBAR < 0 || PAI < 1 || H < 0 || H > 1 || QBAR < 0 || YBAR < 0 + info = 1; +end + +end + + + diff --git a/tests/estimation/method_of_moments/AnScho_MoM.mod b/tests/estimation/method_of_moments/AnScho/AnScho_MoM.mod similarity index 73% rename from tests/estimation/method_of_moments/AnScho_MoM.mod rename to tests/estimation/method_of_moments/AnScho/AnScho_MoM.mod index 83cec9136..3d0e718c9 100644 --- a/tests/estimation/method_of_moments/AnScho_MoM.mod +++ b/tests/estimation/method_of_moments/AnScho/AnScho_MoM.mod @@ -1,8 +1,8 @@ -% DSGE model used in replication files of +% DSGE model used in replication files of % An, Sungbae and Schorfheide, Frank, (2007), Bayesian Analysis of DSGE Models, Econometric Reviews, 26, issue 2-4, p. 113-172. % Adapted by Willi Mutschler (@wmutschl, willi@mutschler.eu) % ========================================================================= -% Copyright (C) 2020 Dynare Team +% Copyright (C) 2020-2021 Dynare Team % % This file is part of Dynare. % @@ -203,28 +203,33 @@ end @#for mommethod in ["GMM", "SMM"] method_of_moments( - % Necessery options - mom_method = @{mommethod} % method of moments method; possible values: GMM|SMM - , datafile = 'AnScho_MoM_data_@{orderApp}.mat' % name of filename with data + % Necessery options + mom_method = @{mommethod} % method of moments method; possible values: GMM|SMM + , datafile = 'AnScho_MoM_data_@{orderApp}.mat' % name of filename with data - % Options for both GMM and SMM + % Options for both GMM and SMM % , bartlett_kernel_lag = 20 % bandwith in optimal weighting matrix , order = @{orderApp} % order of Taylor approximation in perturbation - % , penalized_estimator % use penalized optimization + % , penalized_estimator % include deviation from prior mean as additional moment restriction and use prior precision as weight , pruning % use pruned state space system at higher-order % , verbose % display and store intermediate estimation results - , weighting_matrix = ['optimal'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename - , additional_optimizer_steps = [4] % vector of numbers for the iterations in the 2-step feasible method of moments - % , prefilter=0 % demean each data series by its empirical mean and use centered moments - % - % Options for SMM + , weighting_matrix = ['optimal'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename. Size of cell determines stages in iterated estimation, e.g. two state with ['DIAGONAL','OPTIMAL'] + %, weighting_matrix_scaling_factor=1 % scaling of weighting matrix in objective function + , se_tolx=1e-6 % step size for numerical computation of standard errors + + % Options for SMM + % , burnin=500 % number of periods dropped at beginning of simulation % , bounded_shock_support % trim shocks in simulation to +- 2 stdev - % , drop = 500 % number of periods dropped at beginning of simulation % , seed = 24051986 % seed used in simulations % , simulation_multiple = 5 % multiple of the data length used for simulation - % - % General options - %, dirname = 'MM' % directory in which to store estimation output + + % Options for GMM + @#if mommethod == "GMM" + , analytic_standard_errors % compute standard errors using analytical derivatives + @#endif + + % General options + % , dirname = 'MM' % directory in which to store estimation output % , graph_format = EPS % specify the file format(s) for graphs saved to disk % , nodisplay % do not display the graphs, but still save them to disk % , nograph % do not create graphs (which implies that they are not saved to the disk nor displayed) @@ -232,44 +237,50 @@ end % , plot_priors = 1 % control plotting of priors % , prior_trunc = 1e-10 % probability of extreme values of the prior density that is ignored when computing bounds for the parameters % , TeX % print TeX tables and graphics - % - % Data and model options - %, first_obs = 501 % number of first observation - % , logdata % if loglinear is set, this option is necessary if the user provides data already in logs, otherwise the log transformation will be applied twice (this may result in complex data) - % , loglinear % computes a log-linear approximation of the model instead of a linear approximation - , nobs = 250 % number of observations - % , xls_sheet = willi % name of sheet with data in Excel + + % Data and model options + % , first_obs = 501 % number of first observation + % , logdata % if data is already in logs + , nobs = 250 % number of observations + % , prefilter=0 % demean each data series by its empirical mean and use centered moments + + % , xls_sheet = data % name/number of sheet with data in Excel % , xls_range = B2:D200 % range of data in Excel sheet - % - % Optimization options that can be set by the user in the mod file, otherwise default values are provided - % , analytic_derivation % uses analytic derivatives to compute standard errors for GMM - %, huge_number=1D10 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons - , mode_compute = @{optimizer} % specifies the optimizer for minimization of moments distance, note that by default there is a new optimizer - %, optim = ('TolFun', 1e-5 - % ,'TolX', 1e-6 - % ) % a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute - , silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between - % , tolf = 1e-5 % convergence criterion on function value for numerical differentiation - % , tolx = 1e-6 % convergence criterion on funciton input for numerical differentiation - % - % % Numerical algorithms options + + % Optimization options that can be set by the user in the mod file, otherwise default values are provided + % , huge_number=1e7 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons + , mode_compute = @{optimizer} % specifies the optimizer for minimization of moments distance + , additional_optimizer_steps = [1] % vector of additional mode-finders run after mode_compute + % optim: a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute, some exemplary common options: + , optim = ('TolFun' , 1e-6 % termination tolerance on the function value, a positive scalar + ,'TolX' , 1e-6 % termination tolerance on x, a positive scalar + ,'MaxIter' , 3000 % maximum number of iterations allowed, a positive integer + ,'MaxFunEvals' , 1D6 % maximum number of function evaluations allowed, a positive integer + % ,'UseParallel' , 1 % when true (and supported by optimizer) solver estimates gradients in parallel (using Matlab/Octave's parallel toolbox) + % ,'Jacobian' , 'off' % when 'off' gradient-based solvers approximate Jacobian using finite differences; for GMM we can also pass the analytical Jacobian to gradient-based solvers by setting this 'on' + ) + , silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + + % Numerical algorithms options % , aim_solver % Use AIM algorithm to compute perturbation approximation + % , k_order_solver % use k_order_solver in higher order perturbation approximations % , dr=default % method used to compute the decision rule; possible values are DEFAULT, CYCLE_REDUCTION, LOGARITHMIC_REDUCTION % , dr_cycle_reduction_tol = 1e-7 % convergence criterion used in the cycle reduction algorithm + % , dr_logarithmic_reduction_tol = 1e-12 % convergence criterion used in the logarithmic reduction algorithm % , dr_logarithmic_reduction_maxiter = 100 % maximum number of iterations used in the logarithmic reduction algorithm - % , dr_logarithmic_reduction_tol = 1e-12 % convergence criterion used in the cycle reduction algorithm - % , k_order_solver % use k_order_solver in higher order perturbation approximations % , lyapunov = DEFAULT % algorithm used to solve lyapunov equations; possible values are DEFAULT, FIXED_POINT, DOUBLING, SQUARE_ROOT_SOLVER % , lyapunov_complex_threshold = 1e-15 % complex block threshold for the upper triangular matrix in symmetric Lyapunov equation solver % , lyapunov_fixed_point_tol = 1e-10 % convergence criterion used in the fixed point Lyapunov solver % , lyapunov_doubling_tol = 1e-16 % convergence criterion used in the doubling algorithm % , sylvester = default % algorithm to solve Sylvester equation; possible values are DEFAULT, FIXED_POINT % , sylvester_fixed_point_tol = 1e-12 % convergence criterion used in the fixed point Sylvester solver - % , qz_criterium = 0.999999 % value used to split stable from unstable eigenvalues in reordering the Generalized Schur decomposition used for solving first order problems [IS THIS CORRET @wmutschl] + % , qz_criterium = 0.999999 % value used to split stable from unstable eigenvalues in reordering the Generalized Schur decomposition used for solving first order problems % , qz_zero_threshold = 1e-6 % value used to test if a generalized eigenvalue is 0/0 in the generalized Schur decomposition - @#if mommethod == "GMM" - , analytic_standard_errors - @#endif + % , schur_vec_tol=1e-11 % tolerance level used to find nonstationary variables in Schur decomposition of the transition matrix + % , mode_check % plot the target function for values around the computed minimum for each estimated parameter in turn + % , mode_check_neighbourhood_size = 5 % width of the window (expressed in percentage deviation) around the computed minimum to be displayed on the diagnostic plots + % , mode_check_symmetric_plots=1 % ensure that the check plots are symmetric around the minimum + % , mode_check_number_of_points = 20 % number of points around the minimum where the target function is evaluated (for each parameter) ); @#endfor diff --git a/tests/estimation/method_of_moments/RBC_Andreasen_Data_2.mat b/tests/estimation/method_of_moments/RBC/RBC_Andreasen_Data_2.mat similarity index 100% rename from tests/estimation/method_of_moments/RBC_Andreasen_Data_2.mat rename to tests/estimation/method_of_moments/RBC/RBC_Andreasen_Data_2.mat diff --git a/tests/estimation/method_of_moments/RBC/RBC_MoM_Andreasen.mod b/tests/estimation/method_of_moments/RBC/RBC_MoM_Andreasen.mod new file mode 100644 index 000000000..b9191a1f7 --- /dev/null +++ b/tests/estimation/method_of_moments/RBC/RBC_MoM_Andreasen.mod @@ -0,0 +1,230 @@ +% Tests SMM and GMM routines +% +% Copyright (C) 2020-2021 Dynare Team +% +% This file is part of Dynare. +% +% Dynare is free software: you can redistribute it and/or modify +% it under the terms of the GNU General Public License as published by +% the Free Software Foundation, either version 3 of the License, or +% (at your option) any later version. +% +% Dynare is distributed in the hope that it will be useful, +% but WITHOUT ANY WARRANTY; without even the implied warranty of +% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +% GNU General Public License for more details. +% +% You should have received a copy of the GNU General Public License +% along with Dynare. If not, see . +% ========================================================================= + +% Define testscenario +@#define orderApp = 2 +@#define estimParams = 1 + +% Note that we will set the numerical optimization tolerance levels very large to speed up the testsuite +@#define optimizer = 13 + + +@#include "RBC_MoM_common.inc" + +shocks; +var u_a; stderr 0.0072; +end; + +varobs c iv n; + + +@#if estimParams == 0 +estimated_params; + DELTA, 0.025; + BETTA, 0.984; + B, 0.5; + ETAc, 2; + ALFA, 0.667; + RHOA, 0.979; + stderr u_a, 0.0072; +end; +@#endif + +@#if estimParams == 1 +estimated_params; + DELTA, , 0, 1; + BETTA, , 0, 1; + B, , 0, 1; + ETAc, , 0, 10; + ALFA, , 0, 1; + RHOA, , 0, 1; + stderr u_a, , 0, 1; +end; +@#endif + +@#if estimParams == 2 +estimated_params; + DELTA, 0.025, 0, 1, normal_pdf, 0.02, 0.5; + BETTA, 0.98, 0, 1, beta_pdf, 0.90, 0.25; + B, 0.45, 0, 1, normal_pdf, 0.40, 0.5; + %ETAl, 1, 0, 10, normal_pdf, 0.25, 0.0.1; + ETAc, 1.8, 0, 10, normal_pdf, 1.80, 0.5; + ALFA, 0.65, 0, 1, normal_pdf, 0.60, 0.5; + RHOA, 0.95, 0, 1, normal_pdf, 0.90, 0.5; + stderr u_a, 0.01, 0, 1, normal_pdf, 0.01, 0.5; + %THETA, 3.48, 0, 10, normal_pdf, 0.25, 0.0.1; +end; +@#endif + +% Simulate data +%stoch_simul(order=@{orderApp},pruning,nodisplay,nomoments,periods=500); +%save('RBC_MoM_data_@{orderApp}.mat', options_.varobs{:} ); +%pause(1); + + +estimated_params_init(use_calibration); +end; + + +%-------------------------------------------------------------------------- +% Method of Moments Estimation +%-------------------------------------------------------------------------- +matched_moments; +c; +n; +iv; +c*c; +c*iv; +iv*n; +iv*iv; +n*c; +n*n; +c*c(-1); +n*n(-1); +iv*iv(-1); + +c*c(-3); +n*n(-3); +iv*iv(-3); + +c*c(-5); +n*n(-5); +iv*iv(-5); +end; + +% get indices in declaration order +ic = strmatch('c', M_.endo_names,'exact'); +iiv = strmatch('iv', M_.endo_names,'exact'); +in = strmatch('n', M_.endo_names,'exact'); +% first entry: number of variable in declaration order +% second entry: lag +% third entry: power + +matched_moments_ = { + [ic ] [0 ], [1 ]; + [in ] [0 ], [1 ]; + [iiv ] [0 ], [1 ]; + + [ic ic ] [0 0], [1 1]; + [ic iiv] [0 0], [1 1]; + %[ic in ] [0 0], [1 1]; + %[iiv ic ] [0 0], [1 1]; + [in iiv] [0 0], [1 1]; + [iiv iiv] [0 0], [1 1]; + [ic in] [0 0], [1 1]; + %[in iiv] [0 0], [1 1]; + [in in ] [0 0], [1 1]; + + [ic ic ] [0 -1], [1 1]; + [in in ] [0 -1], [1 1]; + [iiv iiv] [0 -1], [1 1]; + + [ic ic ] [0 -3], [1 1]; + [in in ] [0 -3], [1 1]; + [iiv iiv] [0 -3], [1 1]; + + [ic ic ] [0 -5], [1 1]; + [in in ] [0 -5], [1 1]; + [iiv iiv] [0 -5], [1 1]; + +}; + +if ~isequal(M_.matched_moments,matched_moments_) + error('Translation to matched_moments-block failed') +end + + +method_of_moments( + % Necessery options + mom_method = GMM % method of moments method; possible values: GMM|SMM + , datafile = 'RBC_Andreasen_Data_2.mat' % name of filename with data + + % Options for both GMM and SMM + % , bartlett_kernel_lag = 20 % bandwith in optimal weighting matrix + , order = @{orderApp} % order of Taylor approximation in perturbation + % , penalized_estimator % include deviation from prior mean as additional moment restriction and use prior precision as weight + % , pruning % use pruned state space system at higher-order + % , verbose % display and store intermediate estimation results + , weighting_matrix = ['DIAGONAL','OPTIMAL'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename. Size of cell determines stages in iterated estimation, e.g. two state with ['DIAGONAL','OPTIMAL'] + % , weighting_matrix_scaling_factor=1 % scaling of weighting matrix in objective function + , se_tolx=1e-6 % step size for numerical computation of standard errors + + % Options for SMM + % , burnin=500 % number of periods dropped at beginning of simulation + % , bounded_shock_support % trim shocks in simulation to +- 2 stdev + % , seed = 24051986 % seed used in simulations + % , simulation_multiple = 5 % multiple of the data length used for simulation + + % Options for GMM + % , analytic_standard_errors % compute standard errors using analytical derivatives + + % General options + % , dirname = 'MM' % directory in which to store estimation output + % , graph_format = EPS % specify the file format(s) for graphs saved to disk + % , nodisplay % do not display the graphs, but still save them to disk + % , nograph % do not create graphs (which implies that they are not saved to the disk nor displayed) + % , noprint % do not print stuff to console + % , plot_priors = 1 % control plotting of priors + % , prior_trunc = 1e-10 % probability of extreme values of the prior density that is ignored when computing bounds for the parameters + , TeX % print TeX tables and graphics + + % Data and model options + % , first_obs = 501 % number of first observation + % , logdata % if data is already in logs + % , nobs = 250 % number of observations + % , prefilter=0 % demean each data series by its empirical mean and use centered moments + % , xls_sheet = data % name/number of sheet with data in Excel + % , xls_range = B2:D200 % range of data in Excel sheet + + % Optimization options that can be set by the user in the mod file, otherwise default values are provided + % , huge_number=1e7 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons + , mode_compute = 3 % specifies the optimizer for minimization of moments distance + , additional_optimizer_steps = [13] % vector of additional mode-finders run after mode_compute + % optim: a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute, some exemplary common options: + , optim = ('TolFun' , 1D-6 % termination tolerance on the function value, a positive scalar + ,'TolX' , 1e-6 % termination tolerance on x, a positive scalar + % ,'MaxIter' , 3000 % maximum number of iterations allowed, a positive integer + % ,'MaxFunEvals' , 1D6 % maximum number of function evaluations allowed, a positive integer + % ,'UseParallel' , 1 % when true (and supported by optimizer) solver estimates gradients in parallel (using Matlab/Octave's parallel toolbox) + % ,'Jacobian' , 'off' % when 'off' gradient-based solvers approximate Jacobian using finite differences; for GMM we can also pass the analytical Jacobian to gradient-based solvers by setting this 'on' + ) + % , silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + + % Numerical algorithms options + % , aim_solver % Use AIM algorithm to compute perturbation approximation + % , k_order_solver % use k_order_solver in higher order perturbation approximations + % , dr=default % method used to compute the decision rule; possible values are DEFAULT, CYCLE_REDUCTION, LOGARITHMIC_REDUCTION + % , dr_cycle_reduction_tol = 1e-7 % convergence criterion used in the cycle reduction algorithm + % , dr_logarithmic_reduction_tol = 1e-12 % convergence criterion used in the logarithmic reduction algorithm + % , dr_logarithmic_reduction_maxiter = 100 % maximum number of iterations used in the logarithmic reduction algorithm + % , lyapunov = DEFAULT % algorithm used to solve lyapunov equations; possible values are DEFAULT, FIXED_POINT, DOUBLING, SQUARE_ROOT_SOLVER + % , lyapunov_complex_threshold = 1e-15 % complex block threshold for the upper triangular matrix in symmetric Lyapunov equation solver + % , lyapunov_fixed_point_tol = 1e-10 % convergence criterion used in the fixed point Lyapunov solver + % , lyapunov_doubling_tol = 1e-16 % convergence criterion used in the doubling algorithm + % , sylvester = default % algorithm to solve Sylvester equation; possible values are DEFAULT, FIXED_POINT + % , sylvester_fixed_point_tol = 1e-12 % convergence criterion used in the fixed point Sylvester solver + % , qz_criterium = 0.999999 % value used to split stable from unstable eigenvalues in reordering the Generalized Schur decomposition used for solving first order problems + % , qz_zero_threshold = 1e-6 % value used to test if a generalized eigenvalue is 0/0 in the generalized Schur decomposition + % , schur_vec_tol=1e-11 % tolerance level used to find nonstationary variables in Schur decomposition of the transition matrix + , mode_check % plot the target function for values around the computed minimum for each estimated parameter in turn + % , mode_check_neighbourhood_size = 5 % width of the window (expressed in percentage deviation) around the computed minimum to be displayed on the diagnostic plots + % , mode_check_symmetric_plots=1 % ensure that the check plots are symmetric around the minimum + % , mode_check_number_of_points = 20 % number of points around the minimum where the target function is evaluated (for each parameter) + ); \ No newline at end of file diff --git a/tests/estimation/method_of_moments/RBC_MoM_SMM_ME.mod b/tests/estimation/method_of_moments/RBC/RBC_MoM_SMM_ME.mod similarity index 65% rename from tests/estimation/method_of_moments/RBC_MoM_SMM_ME.mod rename to tests/estimation/method_of_moments/RBC/RBC_MoM_SMM_ME.mod index 964862670..b407cd28a 100644 --- a/tests/estimation/method_of_moments/RBC_MoM_SMM_ME.mod +++ b/tests/estimation/method_of_moments/RBC/RBC_MoM_SMM_ME.mod @@ -1,3 +1,5 @@ +% ========================================================================= +% Copyright (C) 2020-2021 Dynare Team % % This file is part of Dynare. % @@ -137,30 +139,31 @@ end @#for mommethod in ["SMM"] method_of_moments( - % Necessery options - mom_method = @{mommethod} % method of moments method; possible values: GMM|SMM - , datafile = 'RBC_MoM_data_@{orderApp}.mat' % name of filename with data + % Necessery options + mom_method = @{mommethod} % method of moments method; possible values: GMM|SMM + , datafile = 'RBC_MoM_data_@{orderApp}.mat' % name of filename with data - % Options for both GMM and SMM + % Options for both GMM and SMM % , bartlett_kernel_lag = 20 % bandwith in optimal weighting matrix , order = @{orderApp} % order of Taylor approximation in perturbation - % , penalized_estimator % use penalized optimization + % , penalized_estimator % include deviation from prior mean as additional moment restriction and use prior precision as weight , pruning % use pruned state space system at higher-order % , verbose % display and store intermediate estimation results - , weighting_matrix = ['identity_matrix'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename - , weighting_matrix_scaling_factor = 10 - , burnin=250 - %, additional_optimizer_steps = [4] % vector of additional mode-finders run after mode_compute - % , prefilter=0 % demean each data series by its empirical mean and use centered moments - % - % Options for SMM + , weighting_matrix = ['identity_matrix'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename. Size of cell determines stages in iterated estimation, e.g. two state with ['DIAGONAL','OPTIMAL'] + , weighting_matrix_scaling_factor=10 % scaling of weighting matrix in objective function + % , se_tolx=1e-6 % step size for numerical computation of standard errors + + % Options for SMM + , burnin=250 % number of periods dropped at beginning of simulation % , bounded_shock_support % trim shocks in simulation to +- 2 stdev - % , drop = 500 % number of periods dropped at beginning of simulation % , seed = 24051986 % seed used in simulations % , simulation_multiple = 5 % multiple of the data length used for simulation - % - % General options - %, dirname = 'MM' % directory in which to store estimation output + + % Options for GMM + % , analytic_standard_errors % compute standard errors using analytical derivatives + + % General options + % , dirname = 'MM' % directory in which to store estimation output % , graph_format = EPS % specify the file format(s) for graphs saved to disk % , nodisplay % do not display the graphs, but still save them to disk % , nograph % do not create graphs (which implies that they are not saved to the disk nor displayed) @@ -168,41 +171,49 @@ end % , plot_priors = 1 % control plotting of priors % , prior_trunc = 1e-10 % probability of extreme values of the prior density that is ignored when computing bounds for the parameters % , TeX % print TeX tables and graphics - % - % Data and model options - %, first_obs = 501 % number of first observation - % , logdata % if loglinear is set, this option is necessary if the user provides data already in logs, otherwise the log transformation will be applied twice (this may result in complex data) - % , loglinear % computes a log-linear approximation of the model instead of a linear approximation - %, nobs = 500 % number of observations - % , xls_sheet = willi % name of sheet with data in Excel + + % Data and model options + % , first_obs = 501 % number of first observation + % , logdata % if data is already in logs + % , nobs = 250 % number of observations + % , prefilter=0 % demean each data series by its empirical mean and use centered moments + % , xls_sheet = data % name/number of sheet with data in Excel % , xls_range = B2:D200 % range of data in Excel sheet - % - % Optimization options that can be set by the user in the mod file, otherwise default values are provided - % , analytic_derivation % uses analytic derivatives to compute standard errors for GMM - %, huge_number=1D10 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons - , mode_compute = @{optimizer} % specifies the optimizer for minimization of moments distance, note that by default there is a new optimizer - %, optim = ('TolFun', 1e-3 - % ,'TolX', 1e-5 - % ) % a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute - %, silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between - % , tolf = 1e-5 % convergence criterion on function value for numerical differentiation - % , tolx = 1e-6 % convergence criterion on funciton input for numerical differentiation - % - % % Numerical algorithms options + + % Optimization options that can be set by the user in the mod file, otherwise default values are provided + % , huge_number=1e7 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons + , mode_compute = @{optimizer} % specifies the optimizer for minimization of moments distance + %, additional_optimizer_steps = [1 2 3 4] % vector of additional mode-finders run after mode_compute + % optim: a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute, some exemplary common options: + % , optim = ('TolFun' , 1e-6 % termination tolerance on the function value, a positive scalar + % ,'TolX' , 1e-6 % termination tolerance on x, a positive scalar + % ,'MaxIter' , 3000 % maximum number of iterations allowed, a positive integer + % ,'MaxFunEvals' , 1D6 % maximum number of function evaluations allowed, a positive integer + % ,'UseParallel' , 1 % when true (and supported by optimizer) solver estimates gradients in parallel (using Matlab/Octave's parallel toolbox) + % ,'Jacobian' , 'off' % when 'off' gradient-based solvers approximate Jacobian using finite differences; for GMM we can also pass the analytical Jacobian to gradient-based solvers by setting this 'on' + % ) + % , silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + + % Numerical algorithms options % , aim_solver % Use AIM algorithm to compute perturbation approximation + % , k_order_solver % use k_order_solver in higher order perturbation approximations % , dr=default % method used to compute the decision rule; possible values are DEFAULT, CYCLE_REDUCTION, LOGARITHMIC_REDUCTION % , dr_cycle_reduction_tol = 1e-7 % convergence criterion used in the cycle reduction algorithm + % , dr_logarithmic_reduction_tol = 1e-12 % convergence criterion used in the logarithmic reduction algorithm % , dr_logarithmic_reduction_maxiter = 100 % maximum number of iterations used in the logarithmic reduction algorithm - % , dr_logarithmic_reduction_tol = 1e-12 % convergence criterion used in the cycle reduction algorithm - % , k_order_solver % use k_order_solver in higher order perturbation approximations % , lyapunov = DEFAULT % algorithm used to solve lyapunov equations; possible values are DEFAULT, FIXED_POINT, DOUBLING, SQUARE_ROOT_SOLVER % , lyapunov_complex_threshold = 1e-15 % complex block threshold for the upper triangular matrix in symmetric Lyapunov equation solver % , lyapunov_fixed_point_tol = 1e-10 % convergence criterion used in the fixed point Lyapunov solver % , lyapunov_doubling_tol = 1e-16 % convergence criterion used in the doubling algorithm % , sylvester = default % algorithm to solve Sylvester equation; possible values are DEFAULT, FIXED_POINT % , sylvester_fixed_point_tol = 1e-12 % convergence criterion used in the fixed point Sylvester solver - % , qz_criterium = 0.999999 % value used to split stable from unstable eigenvalues in reordering the Generalized Schur decomposition used for solving first order problems [IS THIS CORRET @wmutschl] + % , qz_criterium = 0.999999 % value used to split stable from unstable eigenvalues in reordering the Generalized Schur decomposition used for solving first order problems % , qz_zero_threshold = 1e-6 % value used to test if a generalized eigenvalue is 0/0 in the generalized Schur decomposition + % , schur_vec_tol=1e-11 % tolerance level used to find nonstationary variables in Schur decomposition of the transition matrix + % , mode_check % plot the target function for values around the computed minimum for each estimated parameter in turn + % , mode_check_neighbourhood_size = 5 % width of the window (expressed in percentage deviation) around the computed minimum to be displayed on the diagnostic plots + % , mode_check_symmetric_plots=1 % ensure that the check plots are symmetric around the minimum + % , mode_check_number_of_points = 20 % number of points around the minimum where the target function is evaluated (for each parameter) ); @#endfor diff --git a/tests/estimation/method_of_moments/RBC_MoM_common.inc b/tests/estimation/method_of_moments/RBC/RBC_MoM_common.inc similarity index 71% rename from tests/estimation/method_of_moments/RBC_MoM_common.inc rename to tests/estimation/method_of_moments/RBC/RBC_MoM_common.inc index d480e35c0..330dd2fcf 100644 --- a/tests/estimation/method_of_moments/RBC_MoM_common.inc +++ b/tests/estimation/method_of_moments/RBC/RBC_MoM_common.inc @@ -2,7 +2,23 @@ % Andreasen, Fernández-Villaverde, Rubio-Ramírez (2018): "The Pruned State-Space System for Non-Linear DSGE Models: Theory and Empirical Applications", Review of Economic Studies, 85(1):1-49. % Adapted by Willi Mutschler (@wmutschl, willi@mutschler.eu) % ========================================================================= -% Copyright (C) 2020 Dynare Team +% Copyright (C) 2020-2021 Dynare Team +% +% This file is part of Dynare. +% +% Dynare is free software: you can redistribute it and/or modify +% it under the terms of the GNU General Public License as published by +% the Free Software Foundation, either version 3 of the License, or +% (at your option) any later version. +% +% Dynare is distributed in the hope that it will be useful, +% but WITHOUT ANY WARRANTY; without even the implied warranty of +% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +% GNU General Public License for more details. +% +% You should have received a copy of the GNU General Public License +% along with Dynare. If not, see . +% ========================================================================= var k $K$ c $C$ diff --git a/tests/estimation/method_of_moments/RBC/RBC_MoM_optimizer.mod b/tests/estimation/method_of_moments/RBC/RBC_MoM_optimizer.mod new file mode 100644 index 000000000..08b1ed882 --- /dev/null +++ b/tests/estimation/method_of_moments/RBC/RBC_MoM_optimizer.mod @@ -0,0 +1,146 @@ +% Test optimizers +% +% Copyright (C) 2020-2021 Dynare Team +% +% This file is part of Dynare. +% +% Dynare is free software: you can redistribute it and/or modify +% it under the terms of the GNU General Public License as published by +% the Free Software Foundation, either version 3 of the License, or +% (at your option) any later version. +% +% Dynare is distributed in the hope that it will be useful, +% but WITHOUT ANY WARRANTY; without even the implied warranty of +% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +% GNU General Public License for more details. +% +% You should have received a copy of the GNU General Public License +% along with Dynare. If not, see . +% ========================================================================= +% TO DO +% [ ] fix optimizers 11 and 12; +% note that 12 and 102 require GADS_Toolbox which is not available on servers, but need to be tested locally + +% Define testscenario +@#define orderApp = 2 + +% Note that we will set the numerical optimization tolerance levels very large to speed up the testsuite + +@#include "RBC_MoM_common.inc" + +shocks; +var u_a; stderr 0.0072; +end; + +varobs c iv n; + +%-------------------------------------------------------------------------- +% Method of Moments Estimation +%-------------------------------------------------------------------------- +matched_moments; +c; +n; +iv; +c*c; +c*iv; +iv*n; +iv*iv; +n*c; +n*n; +c*c(-1); +n*n(-1); +iv*iv(-1); +end; + +% reduce options to speed up testsuite +options_.newrat.maxiter = 10; +options_.newrat.tolerance.f = 1e-2; +options_.newrat.tolerance.f_analytic = 1e-2; + +options_.mh_jscale = 0.6; +options_.gmhmaxlik.iterations=1; +options_.gmhmaxlik.number=2000; +options_.gmhmaxlik.nclimb=2000; +options_.gmhmaxlik.nscale=2000; +options_.gmhmaxlik.target=0.5; + +options_.solveopt.MaxIter=300; +options_.solveopt.LBGradientStep=1e-3; +options_.solveopt.TolFun = 1e-3; +options_.solveopt.TolX = 1e-3; +options_.solveopt.TolXConstraint=1e-3; + + +@#for estimParams in [0, 1, 2] + clear estim_params_; + @#if estimParams == 0 + estimated_params; + %DELTA, 0.025; + %BETTA, 0.984; + %B, 0.5; + %ETAc, 2; + ALFA, 0.667; + RHOA, 0.979; + stderr u_a, 0.0072; + end; + @#define OPTIMIZERS = [1, 2, 3, 4, 5, 7, 8, 9, 10, 13, 101] + @#endif + + @#if estimParams == 1 + estimated_params; + %DELTA, , 0, 1; + %BETTA, , 0, 1; + %B, , 0, 1; + %ETAc, , 0, 10; + ALFA, , 0, 1; + RHOA, , 0, 1; + stderr u_a, , 0, 1; + end; + @#define OPTIMIZERS = [1, 2, 3, 4, 7, 8, 9, 10, 13, 101] + @#endif + + @#if estimParams == 2 + estimated_params; + %DELTA, 0.025, 0, 1, normal_pdf, 0.02, 0.5; + %BETTA, 0.98, 0, 1, beta_pdf, 0.90, 0.25; + %B, 0.45, 0, 1, normal_pdf, 0.40, 0.5; + %ETAl, 1, 0, 10, normal_pdf, 0.25, 0.0.1; + %ETAc, 1.8, 0, 10, normal_pdf, 1.80, 0.5; + ALFA, 0.65, 0, 1, normal_pdf, 0.60, 0.5; + RHOA, 0.95, 0, 1, normal_pdf, 0.90, 0.5; + stderr u_a, 0.01, 0, 1, normal_pdf, 0.01, 0.5; + %THETA, 3.48, 0, 10, normal_pdf, 0.25, 0.0.1; + end; + @#define OPTIMIZERS = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 13, 101] + @#endif + + estimated_params_init(use_calibration); + end; + + @#for optimizer in OPTIMIZERS + method_of_moments( + mom_method = GMM % method of moments method; possible values: GMM|SMM + , datafile = 'RBC_Andreasen_Data_2.mat' % name of filename with data + , order = @{orderApp} % order of Taylor approximation in perturbation + , weighting_matrix = ['OPTIMAL'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename. Size of cell determines stages in iterated estimation, e.g. two state with ['DIAGONAL','OPTIMAL'] + , nograph % do not create graphs (which implies that they are not saved to the disk nor displayed) + , mode_compute = @{optimizer} % specifies the optimizer for minimization of moments distance + @#if optimizer == 102 + , optim = ('TolFun' , 1D-3 % termination tolerance on the function value, a positive scalar + ,'MaxIter' , 300 % maximum number of iterations allowed, a positive integer + ,'MaxFunEvals' , 1D3 % maximum number of function evaluations allowed, a positive integer + ) + @#else + , optim = ('TolFun' , 1D-3 % termination tolerance on the function value, a positive scalar + ,'TolX' , 1e-3 % termination tolerance on x, a positive scalar + ,'MaxIter' , 300 % maximum number of iterations allowed, a positive integer + ,'MaxFunEvals' , 1D3 % maximum number of function evaluations allowed, a positive integer + ) + @#endif + %, silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + ); + @#endfor +@#endfor + + + diff --git a/tests/estimation/method_of_moments/RBC_MoM_prefilter.mod b/tests/estimation/method_of_moments/RBC/RBC_MoM_prefilter.mod similarity index 61% rename from tests/estimation/method_of_moments/RBC_MoM_prefilter.mod rename to tests/estimation/method_of_moments/RBC/RBC_MoM_prefilter.mod index 7fb29f8ab..326badcd5 100644 --- a/tests/estimation/method_of_moments/RBC_MoM_prefilter.mod +++ b/tests/estimation/method_of_moments/RBC/RBC_MoM_prefilter.mod @@ -1,6 +1,6 @@ % Tests SMM and GMM routines with prefilter, explicit initialization, and estimated_params_init(use_calibration); % -% Copyright (C) 2020 Dynare Team +% Copyright (C) 2020-2021 Dynare Team % % This file is part of Dynare. % @@ -110,31 +110,31 @@ save('test_matrix.mat','weighting_matrix') @#for mommethod in ["GMM", "SMM"] method_of_moments( - % Necessery options - mom_method = @{mommethod} % method of moments method; possible values: GMM|SMM - , datafile = 'RBC_MoM_data_@{orderApp}.mat' % name of filename with data + % Necessery options + mom_method = @{mommethod} % method of moments method; possible values: GMM|SMM + , datafile = 'RBC_MoM_data_@{orderApp}.mat' % name of filename with data - % Options for both GMM and SMM + % Options for both GMM and SMM % , bartlett_kernel_lag = 20 % bandwith in optimal weighting matrix , order = @{orderApp} % order of Taylor approximation in perturbation - % , penalized_estimator % use penalized optimization + % , penalized_estimator % include deviation from prior mean as additional moment restriction and use prior precision as weight , pruning % use pruned state space system at higher-order % , verbose % display and store intermediate estimation results -% , weighting_matrix = 'test_matrix.mat' % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename - , weighting_matrix =['test_matrix.mat','optimal'] - %, weighting_matrix = optimal % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename - %, additional_optimizer_steps = [4] % vector of additional mode-finders run after mode_compute - , prefilter=1 % demean each data series by its empirical mean and use centered moments - , se_tolx=1e-5 - % - % Options for SMM + , weighting_matrix = ['test_matrix.mat','optimal'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename. Size of cell determines stages in iterated estimation, e.g. two state with ['DIAGONAL','OPTIMAL'] + %, weighting_matrix_scaling_factor=1 % scaling of weighting matrix in objective function + , se_tolx=1e-5 % step size for numerical computation of standard errors + + % Options for SMM + , burnin=500 % number of periods dropped at beginning of simulation % , bounded_shock_support % trim shocks in simulation to +- 2 stdev - , burnin = 500 % number of periods dropped at beginning of simulation % , seed = 24051986 % seed used in simulations % , simulation_multiple = 5 % multiple of the data length used for simulation - % - % General options - %, dirname = 'MM' % directory in which to store estimation output + + % Options for GMM + % , analytic_standard_errors % compute standard errors using analytical derivatives + + % General options + % , dirname = 'MM' % directory in which to store estimation output % , graph_format = EPS % specify the file format(s) for graphs saved to disk % , nodisplay % do not display the graphs, but still save them to disk % , nograph % do not create graphs (which implies that they are not saved to the disk nor displayed) @@ -142,38 +142,49 @@ save('test_matrix.mat','weighting_matrix') % , plot_priors = 1 % control plotting of priors % , prior_trunc = 1e-10 % probability of extreme values of the prior density that is ignored when computing bounds for the parameters % , TeX % print TeX tables and graphics - % - % Data and model options - %, first_obs = 501 % number of first observation - % , logdata % if loglinear is set, this option is necessary if the user provides data already in logs, otherwise the log transformation will be applied twice (this may result in complex data) - % , loglinear % computes a log-linear approximation of the model instead of a linear approximation - %, nobs = 500 % number of observations - % , xls_sheet = willi % name of sheet with data in Excel + + % Data and model options + % , first_obs = 501 % number of first observation + % , logdata % if data is already in logs + , nobs = 250 % number of observations + , prefilter=1 % demean each data series by its empirical mean and use centered moments + + % , xls_sheet = data % name/number of sheet with data in Excel % , xls_range = B2:D200 % range of data in Excel sheet - % - % Optimization options that can be set by the user in the mod file, otherwise default values are provided - % , analytic_derivation % uses analytic derivatives to compute standard errors for GMM - %, huge_number=1D10 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons - , mode_compute = @{optimizer} % specifies the optimizer for minimization of moments distance, note that by default there is a new optimizer - %, optim = ('TolFun', 1e-3 - % ,'TolX', 1e-5 - % ) % a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute - %, silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between - % - % % Numerical algorithms options + + % Optimization options that can be set by the user in the mod file, otherwise default values are provided + % , huge_number=1e7 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons + , mode_compute = @{optimizer} % specifies the optimizer for minimization of moments distance + %, additional_optimizer_steps = [7] % vector of additional mode-finders run after mode_compute + % optim: a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute, some exemplary common options: + % , optim = ('TolFun' , 1e-6 % termination tolerance on the function value, a positive scalar + % ,'TolX' , 1e-6 % termination tolerance on x, a positive scalar + % ,'MaxIter' , 3000 % maximum number of iterations allowed, a positive integer + % ,'MaxFunEvals' , 1D6 % maximum number of function evaluations allowed, a positive integer + % ,'UseParallel' , 1 % when true (and supported by optimizer) solver estimates gradients in parallel (using Matlab/Octave's parallel toolbox) + % ,'Jacobian' , 'off' % when 'off' gradient-based solvers approximate Jacobian using finite differences; for GMM we can also pass the analytical Jacobian to gradient-based solvers by setting this 'on' + % ) + % , silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between + + % Numerical algorithms options % , aim_solver % Use AIM algorithm to compute perturbation approximation + % , k_order_solver % use k_order_solver in higher order perturbation approximations % , dr=default % method used to compute the decision rule; possible values are DEFAULT, CYCLE_REDUCTION, LOGARITHMIC_REDUCTION % , dr_cycle_reduction_tol = 1e-7 % convergence criterion used in the cycle reduction algorithm + % , dr_logarithmic_reduction_tol = 1e-12 % convergence criterion used in the logarithmic reduction algorithm % , dr_logarithmic_reduction_maxiter = 100 % maximum number of iterations used in the logarithmic reduction algorithm - % , dr_logarithmic_reduction_tol = 1e-12 % convergence criterion used in the cycle reduction algorithm - % , k_order_solver % use k_order_solver in higher order perturbation approximations % , lyapunov = DEFAULT % algorithm used to solve lyapunov equations; possible values are DEFAULT, FIXED_POINT, DOUBLING, SQUARE_ROOT_SOLVER % , lyapunov_complex_threshold = 1e-15 % complex block threshold for the upper triangular matrix in symmetric Lyapunov equation solver % , lyapunov_fixed_point_tol = 1e-10 % convergence criterion used in the fixed point Lyapunov solver % , lyapunov_doubling_tol = 1e-16 % convergence criterion used in the doubling algorithm % , sylvester = default % algorithm to solve Sylvester equation; possible values are DEFAULT, FIXED_POINT % , sylvester_fixed_point_tol = 1e-12 % convergence criterion used in the fixed point Sylvester solver - % , qz_criterium = 0.999999 % value used to split stable from unstable eigenvalues in reordering the Generalized Schur decomposition used for solving first order problems [IS THIS CORRET @wmutschl] + % , qz_criterium = 0.999999 % value used to split stable from unstable eigenvalues in reordering the Generalized Schur decomposition used for solving first order problems % , qz_zero_threshold = 1e-6 % value used to test if a generalized eigenvalue is 0/0 in the generalized Schur decomposition + % , schur_vec_tol=1e-11 % tolerance level used to find nonstationary variables in Schur decomposition of the transition matrix + % , mode_check % plot the target function for values around the computed minimum for each estimated parameter in turn + % , mode_check_neighbourhood_size = 5 % width of the window (expressed in percentage deviation) around the computed minimum to be displayed on the diagnostic plots + % , mode_check_symmetric_plots=1 % ensure that the check plots are symmetric around the minimum + % , mode_check_number_of_points = 20 % number of points around the minimum where the target function is evaluated (for each parameter) ); @#endfor \ No newline at end of file diff --git a/tests/estimation/method_of_moments/RBC/RBC_MoM_steady_helper.m b/tests/estimation/method_of_moments/RBC/RBC_MoM_steady_helper.m new file mode 100644 index 000000000..9c43619d7 --- /dev/null +++ b/tests/estimation/method_of_moments/RBC/RBC_MoM_steady_helper.m @@ -0,0 +1,40 @@ +% ========================================================================= +% Copyright (C) 2020-2021 Dynare Team +% +% This file is part of Dynare. +% +% Dynare is free software: you can redistribute it and/or modify +% it under the terms of the GNU General Public License as published by +% the Free Software Foundation, either version 3 of the License, or +% (at your option) any later version. +% +% Dynare is distributed in the hope that it will be useful, +% but WITHOUT ANY WARRANTY; without even the implied warranty of +% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +% GNU General Public License for more details. +% +% You should have received a copy of the GNU General Public License +% along with Dynare. If not, see . +% ========================================================================= +function [N, info]= RBC_MoM_steady_helper(THETA,ETAl,ETAc,BETTA,B,C_O_N,W) +info=0; +if ~isreal(C_O_N) + info=1; + N=NaN; + return; +end +if ETAc == 1 && ETAl == 1 + N = (1-BETTA*B)*(C_O_N*(1-B))^-1*W/THETA/(1+(1-BETTA*B)*(C_O_N*(1-B))^-1*W/THETA); +else + % No closed-form solution use a fixed-point algorithm + N0 = 1/3; + try + [N, ~, exitflag] = fsolve(@(N) THETA*(1-N)^(-ETAl)*N^ETAc - (1-BETTA*B)*(C_O_N*(1-B))^(-ETAc)*W, N0,optimset('Display','off','TolX',1e-12,'TolFun',1e-12)); + if exitflag<1 + info=1; + end + catch + N=NaN; + info=1; + end +end \ No newline at end of file diff --git a/tests/estimation/method_of_moments/RBC_MoM_Andreasen.mod b/tests/estimation/method_of_moments/RBC_MoM_Andreasen.mod deleted file mode 100644 index ae6984005..000000000 --- a/tests/estimation/method_of_moments/RBC_MoM_Andreasen.mod +++ /dev/null @@ -1,227 +0,0 @@ -% Tests SMM and GMM routines -% -% Copyright (C) 2020 Dynare Team -% -% This file is part of Dynare. -% -% Dynare is free software: you can redistribute it and/or modify -% it under the terms of the GNU General Public License as published by -% the Free Software Foundation, either version 3 of the License, or -% (at your option) any later version. -% -% Dynare is distributed in the hope that it will be useful, -% but WITHOUT ANY WARRANTY; without even the implied warranty of -% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -% GNU General Public License for more details. -% -% You should have received a copy of the GNU General Public License -% along with Dynare. If not, see . -% ========================================================================= - -% Define testscenario -@#define orderApp = 2 -@#define estimParams = 1 - -% Note that we will set the numerical optimization tolerance levels very large to speed up the testsuite -@#define optimizer = 13 - - -@#include "RBC_MoM_common.inc" - -shocks; -var u_a; stderr 0.0072; -end; - -varobs c iv n; - - -@#if estimParams == 0 -estimated_params; - DELTA, 0.025; - BETTA, 0.984; - B, 0.5; - ETAc, 2; - ALFA, 0.667; - RHOA, 0.979; - stderr u_a, 0.0072; -end; -@#endif - -@#if estimParams == 1 -estimated_params; - DELTA, , 0, 1; - BETTA, , 0, 1; - B, , 0, 1; - ETAc, , 0, 10; - ALFA, , 0, 1; - RHOA, , 0, 1; - stderr u_a, , 0, 1; -end; -@#endif - -@#if estimParams == 2 -estimated_params; - DELTA, 0.025, 0, 1, normal_pdf, 0.02, 0.5; - BETTA, 0.98, 0, 1, beta_pdf, 0.90, 0.25; - B, 0.45, 0, 1, normal_pdf, 0.40, 0.5; - %ETAl, 1, 0, 10, normal_pdf, 0.25, 0.0.1; - ETAc, 1.8, 0, 10, normal_pdf, 1.80, 0.5; - ALFA, 0.65, 0, 1, normal_pdf, 0.60, 0.5; - RHOA, 0.95, 0, 1, normal_pdf, 0.90, 0.5; - stderr u_a, 0.01, 0, 1, normal_pdf, 0.01, 0.5; - %THETA, 3.48, 0, 10, normal_pdf, 0.25, 0.0.1; -end; -@#endif - -% Simulate data -%stoch_simul(order=@{orderApp},pruning,nodisplay,nomoments,periods=500); -%save('RBC_MoM_data_@{orderApp}.mat', options_.varobs{:} ); -%pause(1); - - -estimated_params_init(use_calibration); -end; - - -%-------------------------------------------------------------------------- -% Method of Moments Estimation -%-------------------------------------------------------------------------- -matched_moments; -c; -n; -iv; -c*c; -c*iv; -iv*n; -iv*iv; -n*c; -n*n; -c*c(-1); -n*n(-1); -iv*iv(-1); - -c*c(-3); -n*n(-3); -iv*iv(-3); - -c*c(-5); -n*n(-5); -iv*iv(-5); -end; - -% get indices in declaration order -ic = strmatch('c', M_.endo_names,'exact'); -iiv = strmatch('iv', M_.endo_names,'exact'); -in = strmatch('n', M_.endo_names,'exact'); -% first entry: number of variable in declaration order -% second entry: lag -% third entry: power - -matched_moments_ = { - [ic ] [0 ], [1 ]; - [in ] [0 ], [1 ]; - [iiv ] [0 ], [1 ]; - - [ic ic ] [0 0], [1 1]; - [ic iiv] [0 0], [1 1]; - %[ic in ] [0 0], [1 1]; - %[iiv ic ] [0 0], [1 1]; - [in iiv] [0 0], [1 1]; - [iiv iiv] [0 0], [1 1]; - [ic in] [0 0], [1 1]; - %[in iiv] [0 0], [1 1]; - [in in ] [0 0], [1 1]; - - [ic ic ] [0 -1], [1 1]; - [in in ] [0 -1], [1 1]; - [iiv iiv] [0 -1], [1 1]; - - [ic ic ] [0 -3], [1 1]; - [in in ] [0 -3], [1 1]; - [iiv iiv] [0 -3], [1 1]; - - [ic ic ] [0 -5], [1 1]; - [in in ] [0 -5], [1 1]; - [iiv iiv] [0 -5], [1 1]; - -}; - -if ~isequal(M_.matched_moments,matched_moments_) - error('Translation to matched_moments-block failed') -end - - - method_of_moments( - % Necessery options - mom_method = GMM % method of moments method; possible values: GMM|SMM - , datafile = 'RBC_Andreasen_Data_2.mat' % name of filename with data - - % Options for both GMM and SMM - %, bartlett_kernel_lag = 20 % bandwith in optimal weighting matrix - , order = @{orderApp} % order of Taylor approximation in perturbation - %, penalized_estimator % use penalized optimization - %, pruning % use pruned state space system at higher-order - %, verbose % display and store intermediate estimation results - , weighting_matrix = ['DIAGONAL','OPTIMAL'] % weighting matrix in moments distance objective function; possible values: OPTIMAL|IDENTITY_MATRIX|DIAGONAL|filename - %, weighting_matrix_scaling_factor=1 - , additional_optimizer_steps = [13] % vector of additional mode-finders run after mode_compute - %, prefilter=0 % demean each data series by its empirical mean and use centered moments - % - % Options for SMM - %, bounded_shock_support % trim shocks in simulation to +- 2 stdev - %, drop = 500 % number of periods dropped at beginning of simulation - %, seed = 24051986 % seed used in simulations - %, simulation_multiple = 5 % multiple of the data length used for simulation - %, burnin = 200 - % - % General options - %, dirname = 'MM' % directory in which to store estimation output - %, graph_format = EPS % specify the file format(s) for graphs saved to disk - %, nodisplay % do not display the graphs, but still save them to disk - %, nograph % do not create graphs (which implies that they are not saved to the disk nor displayed) - %, noprint % do not print stuff to console - %, plot_priors = 1 % control plotting of priors - %, prior_trunc = 1e-10 % probability of extreme values of the prior density that is ignored when computing bounds for the parameters - , TeX % print TeX tables and graphics - % - % Data and model options - %, first_obs = 501 % number of first observation - %, logdata % if loglinear is set, this option is necessary if the user provides data already in logs, otherwise the log transformation will be applied twice (this may result in complex data) - %, loglinear % computes a log-linear approximation of the model instead of a linear approximation - %, nobs = 50 % number of observations - % , xls_sheet = willi % name of sheet with data in Excel - % , xls_range = B2:D200 % range of data in Excel sheet - % - % Optimization options that can be set by the user in the mod file, otherwise default values are provided - %, analytic_derivation % uses analytic derivatives to compute standard errors for GMM - %, huge_number=1D10 % value for replacing the infinite bounds on parameters by finite numbers. Used by some optimizers for numerical reasons - , mode_compute = 13 % specifies the optimizer for minimization of moments distance, note that by default there is a new optimizer - , optim = ('TolFun', 1D-6 - ,'TolX', 1D-6 - ) % a list of NAME and VALUE pairs to set options for the optimization routines. Available options depend on mode_compute - %, silent_optimizer % run minimization of moments distance silently without displaying results or saving files in between - , se_tolx = 1e-6 % convergence criterion on funciton input for numerical differentiation - % - % % Numerical algorithms options - %, aim_solver % Use AIM algorithm to compute perturbation approximation - %, dr=DEFAULT % method used to compute the decision rule; possible values are DEFAULT, CYCLE_REDUCTION, LOGARITHMIC_REDUCTION - %, dr_cycle_reduction_tol = 1e-7 % convergence criterion used in the cycle reduction algorithm - %, dr_logarithmic_reduction_maxiter = 100 % maximum number of iterations used in the logarithmic reduction algorithm - %, dr_logarithmic_reduction_tol = 1e-12 % convergence criterion used in the cycle reduction algorithm - %, k_order_solver % use k_order_solver in higher order perturbation approximations - %, lyapunov = DEFAULT % algorithm used to solve lyapunov equations; possible values are DEFAULT, FIXED_POINT, DOUBLING, SQUARE_ROOT_SOLVER - %, lyapunov_complex_threshold = 1e-15 % complex block threshold for the upper triangular matrix in symmetric Lyapunov equation solver - %, lyapunov_fixed_point_tol = 1e-10 % convergence criterion used in the fixed point Lyapunov solver - %, lyapunov_doubling_tol = 1e-16 % convergence criterion used in the doubling algorithm - %, sylvester = default % algorithm to solve Sylvester equation; possible values are DEFAULT, FIXED_POINT - %, sylvester_fixed_point_tol = 1e-12 % convergence criterion used in the fixed point Sylvester solver - %, qz_criterium = 0.999999 % value used to split stable from unstable eigenvalues in reordering the Generalized Schur decomposition used for solving first order problems [IS THIS CORRET @wmutschl] - %, qz_zero_threshold = 1e-6 % value used to test if a generalized eigenvalue is 0/0 in the generalized Schur decomposition - , mode_check - %, mode_check_neighbourhood_size=0.5 - %, mode_check_symmetric_plots=0 - %, mode_check_number_of_points=25 - ); - - - diff --git a/tests/estimation/method_of_moments/RBC_MoM_steady_helper.m b/tests/estimation/method_of_moments/RBC_MoM_steady_helper.m deleted file mode 100644 index b495e27a1..000000000 --- a/tests/estimation/method_of_moments/RBC_MoM_steady_helper.m +++ /dev/null @@ -1,22 +0,0 @@ -function [N, info]= RBC_MoM_steady_helper(THETA,ETAl,ETAc,BETTA,B,C_O_N,W) -info=0; -if ~isreal(C_O_N) - info=1; - N=NaN; - return; -end -if ETAc == 1 && ETAl == 1 - N = (1-BETTA*B)*(C_O_N*(1-B))^-1*W/THETA/(1+(1-BETTA*B)*(C_O_N*(1-B))^-1*W/THETA); -else - % No closed-form solution use a fixed-point algorithm - N0 = 1/3; - try - [N, ~, exitflag] = fsolve(@(N) THETA*(1-N)^(-ETAl)*N^ETAc - (1-BETTA*B)*(C_O_N*(1-B))^(-ETAc)*W, N0,optimset('Display','off','TolX',1e-12,'TolFun',1e-12)); - if exitflag<1 - info=1; - end - catch - N=NaN; - info=1; - end -end \ No newline at end of file diff --git a/tests/estimation/method_of_moments/RBC_MoM_steadystate.m b/tests/estimation/method_of_moments/RBC_MoM_steadystate.m deleted file mode 100644 index ba4ef9240..000000000 --- a/tests/estimation/method_of_moments/RBC_MoM_steadystate.m +++ /dev/null @@ -1,74 +0,0 @@ -% By Willi Mutschler, September 26, 2016. Email: willi@mutschler.eu -function [ys,params,check] = RBCmodel_steadystate(ys,exo,M_,options_) -%% Step 0: initialize indicator and set options for numerical solver -check = 0; -options = optimset('Display','off','TolX',1e-12,'TolFun',1e-12); -params = M_.params; - -%% Step 1: read out parameters to access them with their name -for ii = 1:M_.param_nbr - eval([ M_.param_names{ii} ' = M_.params(' int2str(ii) ');']); -end - -%% Step 2: Check parameter restrictions -if ETAc*ETAl<1 % parameter violates restriction (here it is artifical) - check=1; %set failure indicator - return; %return without updating steady states -end - -%% Step 3: Enter model equations here -A = 1; -RK = 1/BETTA - (1-DELTA); -K_O_N = (RK/(A*(1-ALFA)))^(-1/ALFA); -if K_O_N <= 0 - check = 1; % set failure indicator - return; % return without updating steady states -end -W = A*ALFA*(K_O_N)^(1-ALFA); -IV_O_N = DELTA*K_O_N; -Y_O_N = A*K_O_N^(1-ALFA); -C_O_N = Y_O_N - IV_O_N; -if C_O_N <= 0 - check = 1; % set failure indicator - return; % return without updating steady states -end - -% The labor level -if ETAc == 1 && ETAl == 1 - N = (1-BETTA*B)*(C_O_N*(1-B))^-1*W/THETA/(1+(1-BETTA*B)*(C_O_N*(1-B))^-1*W/THETA); -else - % No closed-form solution use a fixed-point algorithm - N0 = 1/3; - [N,~,exitflag] = fsolve(@(N) THETA*(1-N)^(-ETAl)*N^ETAc - (1-BETTA*B)*(C_O_N*(1-B))^(-ETAc)*W, N0,options); - if exitflag <= 0 - check = 1; % set failure indicator - return % return without updating steady states - end -end - -C=C_O_N*N; -Y=Y_O_N*N; -IV=IV_O_N*N; -K=K_O_N*N; -LA = (C-B*C)^(-ETAc)-BETTA*B*(C-B*C)^(-ETAc); - -k=log(K); -c=log(C); -a=log(A); -iv=log(IV); -y=log(Y); -la=log(LA); -n=log(N); -rk=log(RK); -w=log(W); -%% Step 4: Update parameters and variables -params=NaN(M_.param_nbr,1); -for iter = 1:M_.param_nbr %update parameters set in the file - eval([ 'params(' num2str(iter) ') = ' M_.param_names{iter} ';' ]) -end - -for ii = 1:M_.orig_endo_nbr %auxiliary variables are set automatically - eval(['ys(' int2str(ii) ') = ' M_.endo_names{ii} ';']); -end - -end