Date:         Wed, 12 Mar 1997 10:35:13 CST
Reply-To:     Chris Rellinger <cjr@KESTNBAUM.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Chris Rellinger <cjr@KESTNBAUM.COM>
Subject:      Tobit model via PROC LIFEREG
Comments: cc: dac@dickens.kestnbaum.com, cas@dickens.kestnbaum.com

     In the SAS/STAT User's Guide, Volume 2, Version 6, 4th Edition,
on Page 1024, it describes how to go about using PROC LIFEREG to do a
tobit model.  I can get this to work great and in fact it sorts my
observations pretty well according to the response variable.  However,
using the output out= option and p= option, I cannot get meaningful
predictions.  They are no where near the actuals in either my modeling
OR validation samples.
     Is there a way to do some type of transform to the output scores
or use some other option to get valid, accurate scores?
 
My code looks like this:
 
proc lifereg;
  model (low, resp) = var1 var2 var3 / d=normal;
  output out=out1 p=pred;
 
     I'll be glad to get any help I can at this point.  I have called
tech support and I am still waiting.  Thanx in advance.