Prev: Slow Launch Time for 9.2 on Windows XP SP3
Next: Obtaining box-cox power via univariate procs
From: Jonas Bilenas on 20 Oct 2009 11:52
Do you get the same error when you code up the model statement in
events/trail format? I don't get an error with code attached here:

data test;
input group event count;
datalines;
0 379 1000
1 376 1000
2 405 1000
3 338 1000
4 338 1000
;;
run;

proc logistic data=test descending;
class group (ref='0')/param=ref;
model event/count=group /risklimits;
run;


On Tue, 20 Oct 2009 09:45:43 -0400, Eric Benchimol
<ericbenchimol(a)ROGERS.COM> wrote:

>Hi all!
>
>I'm running SAS 9.1.3 and I'm getting a strange error.
>
>I'm trying to calculate whether the likelihood of event is different
>between 4 groups, using PROC LOGISTIC. My base code is:
>
>proc logistic data=dataset;
>class group(REF='0');
>model event = group /risklimits;
>run;
>
>When I add the continuous variable to this model, I get the error 'All
>observations have the same response. No statistics are computed.'
>However, when I had dichotomous variables, it works fine. An example is:
>
>proc logistic data=dataset;
>class group(REF='0') dichotomous_var;
>model event = group dichotomous_var / risklimits;
>run;
>
>When I try adding the EXACTONLY option (in case I needed EXACT
>statistics), I get the same error.
>
>The event rate amongst groups is definitely exactly equal. When I run
>a proc freq, I get the following event rates:
>Group 0 - 37.89%
>Group 1 - 37.66%
>Group 2 - 40.47%
>Group 3 - 33.77%
>Group 4 - 33.76%
>(CHI-SQAURE 0.2936)
>
>Any ideas what's going on? Can't find a clear answer online.
>
>Thanks in advance!
 | 
Pages: 1
Prev: Slow Launch Time for 9.2 on Windows XP SP3
Next: Obtaining box-cox power via univariate procs