Date:         Sat, 23 Mar 1996 14:31:37 +0000
Reply-To:     Miss LS Smoljanovic <ehmasls@CS6400.MCC.AC.UK>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Miss LS Smoljanovic <ehmasls@CS6400.MCC.AC.UK>
Subject:      LOGISTIC PROCEDURE?

 Dear SAS-L members,
 
  my data set is described on following way:
 GP - a categorical variable (4 categories 1-2-3-4)
 Gender- (F or M)
 Agegroup- a categorical variable (5 categories 1-2-3-4-5).
 
 GP is a variable which has a value 1 in the case the subjects respond on the first mail, a value 2 if they did not respond on first one, but they responded on the second one, a value 3 if the did not repond on first and second but they did it on third and a vlue 4 if they did not respond.
 
 I would like to build logistic model with GP as a response variable and Gender and Agegroup as the explanatory variables on the following way:
 log(p1/1-p1)=beta0 + beta1 * x1 + beta2 * x2 (where p1 is the probability for positive respond on the first mail)
 
 log(p2/1-p1)=beta01+ beta1 * x1 + beta2 * x2 (where p2 is the probability for
 positive respond on the second mail ; so p2/1-p1 is the conditional probability for positive respond on the second mail given that some people already did positive respond on the first mail).
 .
 .
 .
 My question is : can I use the PROC LOGISTIC for testing the parallel lines assumption (assuming that my categorical variable GP is ordinal) i.e. the mocdel given by:
 g(Pr(Y<=i|x))=(1,x')gamma
 for i=1,...,k response levels of Y (response variable), x is the vector of
 explanatory variables, and gamma is a vector of unknown parameters consisting of an intercept gammai0 (SAS/STAT manual, pn 1090).
 Could you please suggest me something else (proc CATMOD //)?
 Thank you in advance.
                                          Lada