Added example (estimation of single equation by ols).

time-shift
Stéphane Adjemian (Scylla) 2017-07-28 17:52:08 +02:00
parent 949a6211cd
commit dd3c4cd88c
2 changed files with 64 additions and 0 deletions

View File

@ -0,0 +1,4 @@
#!/bin/sh
rm -rf estimation_of_a_single_equation
rm -f *.json *.m *.mat *.log

View File

@ -0,0 +1,60 @@
// --+ options: json=compute +--
// Declare the endogenous variables.
var ffr, unrate, cpi;
// Declare the exogenous variables.
varexo e_ffr, e_unrate, e_cpi;
// Declare the model.
model(linear);
[name='ffr']
ffr = adl(ffr, 'p_ffr_ffr', [1:3]) + adl(unrate, 'p_ffr_unrate', 1) + adl(cpi, 'p_ffr_cpi', [4]) + e_ffr;
[name='unrate']
unrate = adl(unrate, 'p_ffr_unrate', [4 2 5]) + adl(cpi, 'p_unrate_cpi', 6) + e_unrate;
[name='cpi']
cpi = adl(ffr, 'p_cpi_ffr', 2) + adl(cpi, 'p_cpi_cpi', [2]) + e_cpi;
end;
// Set true values for the parameters (randomly, I write directly in M_.params
// because I am too lazy to write all the parameter names).
M_.params = .1*randn(M_.param_nbr, 1);
shocks;
var e_ffr; stderr .01;
var e_unrate; stderr .01;
var e_cpi; stderr .01;
end;
// Create a dataset by simulating the model.
oo_ = simul_backward_model([], 1000, options_, M_, oo_);
// Put all the simulated data in a dseries object.
ds1 = dseries(transpose(oo_.endo_simul), 1900Q1, cellstr(M_.endo_names));
// Select a subsample for estimation
ds1 = ds1(1951Q1:1990Q1);
// Print true values of the parameters (for comparison with the estimation results).
dyn_table('True values for the parameters (DGP)', {}, {}, cellstr(M_.param_names), ...
{'Parameter Value'}, 4, M_.params);
/* Run the estimation of equation ffr and display the results. Note that the corresponding
** values in M_.params will be updated.
**
** estimate command
** ----------------
**
** First argument is used to specify the estimation method
** Second argument is used to specify the dataset (an existing dseries object).
** Third argument (and following) is(are) the name(s) of equation(s) to be estimated.
*/
estimate method(ols) data(ds1) ffr;