Simulating from graphical models for ordinal categorical data

Multivariate ordinal categorical data is encountered in many fields of research. For analysis and data reduction the conditional independence properties of these data are studied in graphical models. However, to simulate multivariate ordinal data with a specific conditional independence structure, for use in simulation studies or computer intensive methods of inference, is non-trivial. We present a procedure to simulate data from a graphical model determined by the values of the partial gamma coefficients corresponding to the conditional dependencies in the graphical model. Implementation of the procedure is done in a standalone Pascal program, and as an R function.