The PLM procedure, unlike most SAS/STAT procedures, does not operate primarily on an input data set. Instead, the procedure requires you to specify an item store with the SOURCE= option in the PROC PLM statement. The item store contains the necessary information and context about the statistical model that was fit when the store was created. SAS data sets are used only to provide input information in some circumstances. For example, when scoring a data set or when computing least squares means with specially defined population margins. In other words, instead of reading raw data and fitting a model, the PLM procedure reads the results of a model having been fit.

In order to interact with the item store and to reveal its contents, the PLM procedure supports the SHOW statement which converts item store information into standard ODS tables for viewing and further processing.

The PLM procedure is sensitive to the contents of the item store. For example, if a BAYES statement was in effect when the item store was created, the posterior parameter estimates are saved to the item store so that the PLM procedure can perform postprocessing tasks by taking the posterior distribution of estimable functions into account. As another example, for item stores that are generated by a mixed model procedure using the Satterthwaite or Kenward-Roger (Kenward and Roger 1997) degrees-of-freedom method, these methods continue to be available when the item store contents are processed with the PLM procedure.

Because the PLM procedure does not read data and does not fit a model, the processing time of this procedure is usually considerably less than the processing time of the procedure that generates the item store.