Getting Odds Ratio in PROC GLIMMIX with spline function

[Marc-Flo]

Hello.

I'm currently trying to fit a GLMM for a longitudinal analysis.

In this analysis the dependent variable is binary and is measured several times per subject. We've noticed that the evolution of the response of this dependent variable is clearly not linear over time. In order to take this phenomenon into account, we've insert a spline function over the time covariate. 

At present the initial model look like this:

PROC GLIMMIX DATA = data METHOD = LAPLACE;
CLASS ind profil;
EFFECT spline_time=SPLINE(time / KNOTMETHOD=PERCENTILES(9) DEGREE = 1);
MODEL var_dep(EVENT='1') = spline_time|profil / DIST=BIN LINK = LOGIT;
RANDOM INT time / SUBJECT = ind;
RUN;

The problem with the spline function is that I can't get any odds ratio, even for covariates without spline function.

Is there any way (maybe a macro) to display the odds ratio ?

Thank you.

[SteveDenham]

So long as you are not looking for differences in probabilities (search for the Marginals Macro for this), you can get the odds ratio by exponentiating an appropriate ESTIMATE statement.  Examples can be found at the following links:

https://documentation.sas.com/?docsetId=statug&docsetTarget=statug_glimmix_examples20.htm&docsetVers...  and

https://documentation.sas.com/?docsetId=statug&docsetTarget=statug_plm_examples06.htm&docsetVersion=...

Give the basics for the endpoints mentioned. It should not be too difficult to adapt these to your work.

SteveDenham