Bayesian Loglinear Models: Bayes Factor
You can specify the model assumed for the observed data (Poisson, Multinomial, or Nonparametric). Multinomial distribution is the setting by default. The following options are available only when either the Estimate Bayes Factor or Use Both Methods Bayesian Analysis option is selected.
- Poisson Model
- When selected, the Poisson model is assumed for the observed data.
- Multinomial Model
- When selected, the Multinomial model is assumed for the observed data. This is the default setting.
- Fixed Margins
- Select Grand Total, Row Sum, or Column Sum to specify the fixed marginal totals for the contingency table. Grand Total is the default setting.
- Prior Distribution
- Specify the prior distribution type when estimating the Bayes factor.
- Conjugate
- Select to specify a conjugate prior distribution. Use the Shape
Parameters table to specify the shape parameters ars for Gamma
distribution. You must specify the shape parameters when Conjugate is
selected as the prior distribution type.
When a single value is specified, all ars's are assumed to be equal to this value. ɑrs = 1 is the default setting. If you need to specify more than one value, you can separate the values with blank spaces.
The number of numerical values that are specified in each row and each column must match the dimension of the contingency table. All of the specified values must be > 0.
Click Reset to clear the values.
- Scale Parameter
- Specify the scale parameter b for Gamma distribution. You must specify a single value > 0.
- Mixture Dirichlet
- Select to specify a mixture Dirichlet prior distribution.
- Intrinsic
- Select to specify an intrinsic prior distribution.
- Nonparametric Model
- When selected, the Nonparametric model is assumed for the observed data.
- Fixed Margins
- Select Row Sum or Column Sum to specify the fixed marginal totals for the contingency table. Row Sum is the default setting.
- Prior Distribution
- Specify the parameters for the Dirichlet priors. You must specify the Prior
Distribution parameters when Nonparametric Model is selected.
When a single value is specified, all λs are assumed to be equal to this value.
λs = 1 is the setting by default. If you need to specify more than one value, you
can separate the values with blank spaces. All of the specified values must be > 0. The number of
specified, numerical values must match the dimension of the row or column that is not fixed for the
contingency table.
Click Reset to clear the values.