correlated measurement errors. Tests show that there is a "significant" discrepancy between the
univariate filter and the standard filter in presence of correlated measurement errors...
git-svn-id: https://www.dynare.org/svn/dynare/dynare_v4@2185 ac1d8469-bf42-47a9-8791-bf33cf982152
* Changed header of missing_observations_kalman_filter
* Added the univariate approach. This file handles the cases with measurement errors and/or missing data.
git-svn-id: https://www.dynare.org/svn/dynare/dynare_v4@2160 ac1d8469-bf42-47a9-8791-bf33cf982152
* Added a new tolerance parameter specific to the iteration on the riccati equation.
* Added a kalman filter routine allowing for missing observations.
* I do not distinguish anymore models with and without measurement errors (the same m file is used for both models to evaluate the likelihood). For a model without measurement errors H hat to be set to 0 scalar.
git-svn-id: https://www.dynare.org/svn/dynare/dynare_v4@2148 ac1d8469-bf42-47a9-8791-bf33cf982152