Logistic regression - "unique profiles"

Aleksandr — Wed, 27 Apr 2016 13:39:32 GMT

I am doing a Logistic regression on survival status (0=alive,1= dead) with 'proc logistic' on two categorial variables : urgency (0 = No , 1 = yes ), conscious(0=awake , 1=unconscious, 2 =coma). The data is in table form; an observation for every row (total 200 rows) eg:

1 1 2

0 1 1

However, I am not getting it right because from the output it says 5 unique profiles which can't be right since there should be 2 x 3 =6 unique profiles : two choices for the variable 'urgency' and three choises for the variable 'conscious'. When i do the logistic regression on survival status on only one of these variables then it gives 2 and 3 unique profiles, which i would expect.

Can someone explain why I am not getting it right:

proc logistic data =mydata;
model  surv = urgency  conscious /scale=none aggregate;
run;

Re: Logistic regression - "unique profiles"

Rick_SAS — Sat, 16 Apr 2016 12:28:16 GMT

I don't think there is anything wrong with your syntax, I think it is your data. Although you have two levels of urgency and three levels of conscious, that doesn't imply that you have all six JOINT levels. Run PROC FREQ on your data and I suspect you will discover that you only have five unique combinations of (urgency, conscious):

proc freq data=mydata;
   tables urgency*conscious;
run;

topic Re: Logistic regression - "unique profiles" in Statistical Procedures

Logistic regression - "unique profiles"

Re: Logistic regression - "unique profiles"