• | Marginal likelihood algorithm: Choose which algorithm to use for computing the marginal likelihood after the posterior sampler has finished. The choices are: (1) All; (2) modified harmonic mean with a truncated normal distribution as weighting function; (3) version of the modified harmonic mean with a truncated elliptical distribution as weighting function; (4) ; and (5) None. |
• | Sequential estimation (StartIteration StepLength): Determines how sequential estimation is performed. The StartIteration variable gives the initial sample length, while the StepLength variable determines how many draws are added relative to the previous sample. These variables are used for sequential estimation of the marginal likelihood, and of sequential estimation of the posterior mean and median. |
• | Coverage probability (Start:Increment:End): The coverage probability variable is needed when computing Geweke's modified harmonic mean estimator of the marginal likelihood. The Start variable determines the first probability that is used, the Increment determines what is added to the previous probability value, and the End variable determines the last probability value which is used. |
• | Compute the marginal likelihood sequentially: If this option is check marked, YADA will compute the marginal likelihood sequentially once the posterior sampler has finished drawing parameters from the posterior. Sequential estimation of the marginal likelihood can also be performed at any time from the View menu. |
• | Overwrite old draws: This option determines if old draws will be overwritten on the hard disk or not. If the option is not check marked, then given that old draws exist they will be loaded when running the posterior sampler. If old draws do not exist and the number of parallel chains is at least 2, then draws from the fewer number of chains can optionally be used. In contrast, when the option is check marked, then old draws from fewer total number of posterior draws can optionally be loaded by YADA, while old draws with the same number of posterior draws will be overwritten. |

Additional Information
• | A more detailed description about how to compute the marginal likelihood is found in Section 10 of the YADA Manual. |
Page url:
http://www.texlips.net/yada/help/index.html?dsge_posterior_sampling.htm