What’s New in SAS/ETS 13.2


SEVERITY Procedure

The following features have been added to the SEVERITY procedure :

  • The SEVERITY procedure now supports the CLASS statement and specification of a rich set of regression effects in the SCALEMODEL statement. These include singleton continuous effects, polynomial continuous effects, main CLASS variable effects, interaction effects, nested effects, continuous-by-class effects, continuous-nesting-class effects, and a general combination of the preceding effects.

    Note that when you specify regression effects that contain interactions or CLASS variables, the OUTSCORELIB statement and the INEST= option are not supported.

  • The SEVERITY procedure now supports a new method of saving the estimation results and using them for parameter initialization. The new OUTSTORE= option creates an item store, which is a binary file in a format that is specific to the SEVERITY and HPSEVERITY procedures. You can use an OUTSTORE= item store that is created in one PROC SEVERITY or PROC HPSEVERITY step to initialize the parameters in a subsequent PROC SEVERITY or PROC HPSEVERITY step by specifying the new INSTORE= option. These options are required if you want to save and initialize a scale regression model that contains interaction effects or effects with CLASS variables.

    Both the OUTSTORE= and INSTORE= options are experimental in this release of the SEVERITY procedure.

  • The OUTSCORELIB statement to create scoring functions, which was introduced in the previous version of PROC SEVERITY, is now at production status. Note that the scoring functions are not supported when you specify a scale regression model that contains interaction effects or effects with CLASS variables.

  • The SEVERITY procedure now supports the INITSAMPLE option to limit the number of observations that are used to prepare the empirical distribution function (EDF) estimates. It enables you to speed up the EDF estimation step for large data sets, especially when you specify censoring or truncation effects.

    If you do not specify the INITSAMPLE option, then a uniform random sample of at most 10,000 observations is used for EDF estimation.