topic Re: Proc logistic in Statistical Procedures

Proc logistic

Frank_P — Mon, 13 Dec 2010 10:39:12 GMT

Hi, could anyone explain to me why the classification results of the ctable statement in proc logistic are different from the ones that I calculate myself based on the outputted probability? The values are very similar, but I do not know why they are not equal? I use the greater equal as should be used for the ctable. Could it be related to the rounding?
Thanks for helping. Regards, Frank

Re: Proc logistic

lvm — Tue, 14 Dec 2010 20:03:48 GMT

According the the User's Guide, the probabilities obtained in the output window from the ctable option are rounded to the nearest 0.02. This would likely explain the differences you find for classification. Also, there is rounding when using the outroc= option (but different rounding than for the ctable output).
LVM

Re: Proc logistic

Frank_P — Thu, 16 Dec 2010 07:51:24 GMT

Thanks, that might indeed explain the differences. I could not spot that information myself, thanks again! Regards, Frank

Re: Proc logistic

StatDave — Mon, 03 Jan 2011 22:11:43 GMT

As mentioned in this usage note:

http://support.sas.com/kb/22596

the CTABLE results are based on bias-adjusted predicted probabilities not on the simple, unadjusted probabilities.