Programming the statistical procedures from SAS

how do I run repeated measurement design for multinominal GLMMIX

Frequent Learner
Posts: 1

how do I run repeated measurement design for multinominal GLMMIX

Hi, my study is repeated design, each subject finish three conditions (1,2,3),and the dependent variable is multinomial (not ordinal )(0,1,2), I want to know when subject ,in all three kinds of condition, would more likley to choose  which kinds of options (0,1,2). so I used GLMMIX, not repeated ANOVA. I have some questions about my code. This is my code:


proc glimmix data=behavior method=laplace;
class ID condition grade;
model D_Allocation= condition / dist=multi link=glogit cl ddfm=betwithin oddsratio solution;
random intercept condition /subject=ID type=cv cl;
output pred=p resid=r;


there is warning:
ERROR: Nominal models require that the response variable is a group effect on RANDOM statements. You need to add
question1: When I deletd the link code, it worked out, but it is said "Estimated G matrix is not positive definite". And I want to specify glogit, because it is not ordinal.
question 2: I am confused about  the output for the fixed solution. The dependent variable is three options(0,1,2), predict vairable is condition (1,2,3), but how to understand which condition will be more likley to fall into which category of dependent variable.  the output looks like:
Screen Shot 2017-05-29 at 9.15.00 AM.png
question 3: this is repeated design, each subject have three line data (conditon=1,2,3), I wonder whether it is correct for my random code to specify the repeated data. I am not sure whether I need more code for residue, or other covatrix to specify the repeated measurement.
question 4: I read code quide for degree. I wonder whether it is necessary to write "ddfm=betwithin", considering repeated design.
I hope my question is clear. Thanks so much!!! I have hard time for this in GLMMIX 
Ask a Question
Discussion stats
  • 0 replies
  • 1 in conversation