CSSM Procedure

The simulation smoothing process in PROC CSSM enables you to organize the simulation output in three tables: the OUTSTATE= table, which contains the simulations of the latent states; the OUTCOMP= table, which contains the simulations of the model components; and the OUTPRED= table, which contains the simulations of the predictions of the missing response variable values. All these tables share three columns: the BY statement variables column; the SEED variable column, which contains the seed value that is associated with the random draw; and the ID statement variable column. The values of the SEED variable range from integer to integer+nsim–1, where integer is the value of the SEED= option and nsim is the number of draws that you specify in the NSIM= option. Additionally, the mean of the sampling distribution for each simulated quantity is also included for reference, and the SEED value that is assigned to it is always –1. The mean of the sampling distribution is computed in the smoothing phase (for more information, see the section Smoothing Phase). The other columns in these tables depend on the specified model. The details are presented in the following sections.