Run the posterior mode estimation routine for the VAR model. The estimation procedure is iterative using the maximum likelihood estimates of the parameters to create starting values. Each iteration now involves three steps. First, the steady-state parameters are calculated conditional on the parameters on lags and the covariance matrix of the residuals using the first order condition from the log posterior for this parameter group. Second, the parameters on lags are computed conditional on the other parameters using the first order condition for this parameter group. Finally, the covariance matrix is computed in the same way. Once, a new set of parameters is available, the log posterior is computed and its values is compared with the previous value. If the absolute change in the log posterior is sufficiently small the algorithms has converged to a local maximum.
The maximum number of iterations and the tolerance level that the posterior mode estimation algorithms uses can be chosen in the Optimization frame on the Options tab.
Additional Information
• | A detailed description about posterior mode estimation in the Bayesian VAR model can be found in Section 14.2 of the YADA Manual. |
Page url: http://www.texlips.net/yada/help/index.html?estimate_posterior_mode2.htm