The Monte Carlo filtering function relies on a large number of draws from the prior distribution. For each draw the DSGE model solution properties are checked and organized into three categories: (1) a unique and stable (convergent) solution; (2) indeterminacy; and (3) no stable solution.
A pre-requisite for Monte Carlo filtering is that the model does not involve a system prior. The latter type of prior forces the prior distribution to be based on unique and convergent parameter values.
The function computes Kolmogorov-Smirnov tests for each estimated parameter of the hypothesis that the distributions are the same for case (1) and cases (2) & (3). It is also possible to scatter plot pairs of parameters with each dot colored by its model solution property. The Monte Carlo filtering approach has been suggested in Ratto (2008).
Additional Information
• | A more detailed description about the Monte Carlo filtering can also be found in Section 11.14 of the YADA Manual. |
Page url: http://www.texlips.net/yada/help/index.html?monte_carlo_filtering.htm