Overview (BAYES REGRESSION command)

Options

Analysis options

You can specify the significance level for computing credible intervals, the tolerance value for numerical methods, and the maximum number of iterations the AGL method can go through in its computations.

Model design

You can specify the model design for the analysis, which includes regression weights, categorical variables, scale variables

Bayesian analysis method

You can specify POSTERIOR, BAYESFACTOR, or BOTH.

Bayes factor settings

You can specify the approach that is used to estimate the Bayes factor.

Prior distribution settings

You can specify REFERENCE and CONJUGATE prior distribution settings, as well as define shape parameters, scale parameters, mean vector, and lower triangle values in the variance-covariance matrix for the multivariate normal prior.

Regressors

You can specify an observed vector with the values for the regressors.

Statistics to be scored

You can specify the statistics to be scored for the Bayesian prediction distribution.

PMML file generation

You can specify a target file to generate the PMML file.

Output plots

You can control the plots that are output including:

• Covariates to be plotted
• Factors to be plotted
• Maximum factor levels to be plotted
• Plot intercept term
• Plot variance of errors
• Plot the predictive distribution

F-tests

You can invoke (partial) F-tests.

Basic specification

• BAYES REGRESSION dependent_varname [BY factor_list] [WITH covariate_list] invokes the linear regression procedure, and defines a full model.

Subcommand order

• The subcommands can be named in any order.

Syntax rules

• Only one regression weight variable is allowed.
• The specified FACTORS must be a subset of factor_list.
• The specified COVARIATES must be a subset of covariate_list.
• BAYESPRED can be TRUE if values are specified for REGRESSORS.
• VARIABLES must be a subset of COVARIATES and FACTORS.
• INTERCEPT must be TRUE when VARIABLES is empty.