Date:         Fri, 23 Sep 2011 16:13:43 -0400
Reply-To:     Santosh Nazare <santosh.nazare@MAINE.GOV>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Santosh Nazare <santosh.nazare@MAINE.GOV>
Subject:      Logistic regression of grouped data

Hi SAS experts,

I am trying to analyze the aggregated data with logistic regression. I have
three four independent variables (facility, race, age_group and sex) with
outcome of positive test result.

My data is aggregated on each of these independent variables
e.g.
Variables Total test   Positive result
Male       908          26
Female     325          12

Whites     600           4
Blacks     800           7    etc. etc.


I have similar data for two facilities. These facilities differ on education
program. My hypothesis is to test whether there is difference in number of
positive test results in two facilities.

So far, I tried following model, where I can only use one independent
variable in addition to facility.


Data one;
Input sex total positive facility;
cards;
0 908 26 1
1 325 12 1
0 3101 23 0
1 937 9 0
;
run;


proc logistic data=one;
  model positive/total=race facility;
  run;


Any suggestions on how to rearrange the data to model other independent
variables too in this model?

Note: I cannot get other data, I have to use what I have with me.


Thank you,

-San