Date:         Sun, 5 Oct 2003 20:05:56 GMT
Reply-To:     julierog@ix.netcom.com
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Roger Lustig <trovato@VERIZON.NET>
Subject:      Re: proc logistic query
Content-Type: text/plain; charset=us-ascii; format=flowed

Dear Rob:
Check out

http://ftp.sas.com/techsup/download/stat/jackboot.html

which contains macros that will do bootstrap & jackknife resampling for
most regression procedures.

Best,

Roger

Rob Appleby wrote:
> Hello I have what I am sure is a rather inane query relating to proc
> logistic, specifically, how to create a resampling program/macro
> involving a training and testing data set? Please find the program for
> the train/test phases below, which appears to work fine. What I would
> like to do now is reiterate the process (say 500 times) so that
> essentially the the training sets are different each time (and
> therefore the predicted probabilities inputted into the test phase)
> and then obtain an overall summary of misclassification and various
> other output statistics.
>
> The data set I am using is relatively small (n = 441) and I worry
> about the spectrum of 'goodness of fit'and maximum likelihood results
> I get each time I run the procedure.
> Any suggestions and advice would be very much appreciated
> Thanks for your time,
> Rob
>
> DATA FItrain FItest;
> SET Fraser2;
> IF RANUNI(0) < = 1/2 then OUTPUT  FItrain;
> ELSE OUTPUT FItest;
> RUN;
> title 'Severe Incidents Model';
> proc logistic data = FItrain order = data outest = PARMS descending
> simple ;
> class Walking (ref = FIRST) TOD jmale (ref =
> FIRST) admale (ref = FIRST) submale (ref = FIRST) season;
> model absolute = season TOD Walking
> admale submale jmale / rsq lackfit EXPB SELECTION =
> FORWARD SLE = 0.20 SLS = 0.10 CTABLE;
> output out=out1 p=p;
> run;
> proc logistic data = FItest order = data inest = PARMS descending
> simple ;
> class Walking (ref = FIRST) TOD jmale (ref =
> FIRST) admale (ref = FIRST) submale (ref = FIRST) season;
> model absolute = season TOD Walking
> admale submale jmale / rsq lackfit maxiter = 0
> EXPB;
> output out=out2 p=p;
> run;
> proc print data=out2;