topic Re: Multiple binary outcomes in Statistical Procedures

Multiple binary outcomes

kaushal2040 — Tue, 28 Oct 2014 16:29:08 GMT

Hello Forum,

I was wondering if there is way to model multiple binary outcomes in SAS?

Thanks !!!

Re: Multiple binary outcomes

ballardw — Tue, 28 Oct 2014 18:02:26 GMT

Your question is kind of sketchy. You may get more useable responses you provide some example data and what your are trying to do.

Re: Multiple binary outcomes

kaushal2040 — Tue, 28 Oct 2014 18:36:51 GMT

Hi ballardw,

In MANOVA (Multivariate Analysis of Variance) we can fit multiple dependent variables but they have to be continuous . Here I am trying to model multiple (alcohol, smoking, drug) binary (0/1) outcome with some set of predictors . Proc Logistic can handle only one dependent variable. I don't know if there is multivariate counterpart of proc logistic.

Here is mock data.

DATA;

INPUT ALCOHOL SMOKE DRUG GENDER $ RACE $;

CARDS;

1 0 1 MALE WHITE

0 1 0 FEMALE HISPANIC

1 0 1 MALE BLACK

...................

..................

;

RUN;

Thanks !!!

Re: Multiple binary outcomes

SteveDenham — Wed, 29 Oct 2014 19:38:33 GMT

PROC CATMOD presents some possibilities. Start with Example 32.4 Log-Linear Model, Three Dependent Variables, and work up from there.

Steve Denham

Re: Multiple binary outcomes

StatDave — Mon, 03 Nov 2014 19:24:22 GMT

See the "Multivariate logistic model" item the following note:

http://support.sas.com/kb/22871

Not quite the log-linear model that Steve mentioned, you could use CATMOD with the RESPONSE LOGITS; statement when specifying multiple response variables. The note also mentions QLIM for the bivariate case.

Re: Multiple binary outcomes

kaushal2040 — Tue, 04 Nov 2014 18:19:22 GMT

Thanks, SteveDenham and StatDave@sas