topic How to generate predicted values from PROC SEVERITY? in SAS Forecasting and Econometrics

How to generate predicted values from PROC SEVERITY?

topkatz — Tue, 15 Jul 2014 22:12:13 GMT

This is a very basic question. Most of the SAS statistical and econometric modeling PROCs have some kind of option to produce an output data set with predicted values. PROC SEVERITY does not appear to me to provide such an option. Is this true? How can I generate predicted values from a PROC SEVERITY model? Can it be done via PROC PLM (in SAS/STAT)? I see that SAS Model Manager has an autocall macro, %MM_Severity_Create_Scorecode, but I don't have SAS Model Manager. Some PROC SEVERITY models can be replicated easily in PROC GENMOD or PROC GLIMMIX, but PROC SEVERITY also includes some distributions that are not easily available in PROC GENMOD or PROC GLIMMIX (e.g., Weibull).

Thanks!

-- TMK --

20140715

Re: How to generate predicted values from PROC SEVERITY?

ballardw — Tue, 15 Jul 2014 23:01:28 GMT

I suspect the idea would be to use the OUTEST data set, merge back with your raw data and use the parameters estimated to estimate the specific outcome.

Re: How to generate predicted values from PROC SEVERITY?

SteveDenham — Wed, 16 Jul 2014 12:14:53 GMT

What version of SAS/ETS are you on? If it is 13.1, check out the use of the OUTSCORELIB statement (see Example 23.8). It involves PROC FCMP, but the example provides sample code, so it should be possible. Otherwise, I think 's approach will be as good as can be done.

Steve Denham