The file with prior distribution data must be given by either a Lotus 1-2-3 spreadsheet (file extension .wk1) or and Excel spreadsheet (extension .xls). The file needs to be specified with certain required headers and possible values.
The general structure of the prior distribution file shown in Table 6 below, with the 7 required headers first, and the two optional headers last. The headers have been placed row-wise in the Table for expositional reasons. In the prior distribution file they must appear in the same row.
|
||||||||||||||||||
Table 6: The headers and data types for the prior distribution file. |
The upper bound header and the prior parameter 3 header are optional. The upper bound header is only used for parameters with a beta prior. If this header is missing, YADA assumes that the upper bound is 1, while the lower bound is set to 0, i.e. a standardized beta. In that case the value from the lower bound column is ignored. The prior parameter 3 header is used by the Student-t, the logistic (Type I generalized logistic), and the Pareto prior. For the Student-t the value is the number of degrees of freedom, for the logistic it is the shape parameter (which determines if the distribution is left-skewed, symmetric, or right-skewed), and for the Pareto its value determines the origin parameter. If this header is missing or the value is missing or invalid, then the default values given in Table 6 are imposed. These values have been selected such that the Student-t is a Cauchy, the logistic is symmetric, and the Pareto has 0 as the origin (with prior parameter 2 giving the mode and the lower bound).
You can check the most important information that YADA can retrieve about the parameters to be estimated from your selected prior distribution file. Simply, run the Prior Distribution Information function from the View menu.
NOTE: The chi-square, exponential and Erlang distributions are special cases of the gamma and are therefore directly supported in YADA. See Section 2.2.3 of the YADA Manual for details.
NOTE: By combining the file with parameters to update with the gamma prior, you can also allow for parameters that have a Weibull prior distribution. Similarly, through such parameter functions you can also allow for the log-normal and multivariate normal prior via the univariate normal prior for auxiliary parameters, the inverted Wishart via univariate normal and inverted Gamma prior for auxiliary parameters, and the Snedecor (or F) and Dirichlet (multivariate beta) prior through the beta prior distribution.
Additional Information
• | A description of the prior distributions that YADA support can be found in Section 4.2 of the YADA Manual. |
• | A more detailed description about how to setup the prior distribution file can also be found in Section 17.2 of the YADA Manual. |
• | An example file is located in the directory "example\AnSchorfheide\data" directly below the base directory for YADA. The file is simply called "AnSchorfheidePrior.wk1". The same data is also available in the Excel spreadsheet "AnSchorfheidePrior.xls". |
• | Further example files can be found in the "LubikSchorfheide\data" and "SmetsWoutersAER\data" directories in the "example" directory. |
Page url: http://www.texlips.net/yada/help/index.html?prior_distribution_file.htm