Programming the statistical procedures from SAS

Proc Glimmix Odds Ratio

Reply
N/A
Posts: 1

Proc Glimmix Odds Ratio

Hello,

I am using proc glimmix for logistic regression of longitudinal data with dichotonomous outcomes.

The code i am using is this.

proc glimmix data = t1 method = quad;

class Id month feed;

model case = feed month feed*month / s dist = binomial link = logit oddsratio;

random intercept / subject = Id type = un;

run;

The exp(estimates) does not match the odds ratio in the odds ratio table. I wan to know if I am interpreting the tables correctly or am I missing something. Any help in this matter is greatly appreciated. Thx, N

I have attached the out put as a text file "GLIMMIX - odds ratio.txt".

Attachment
New Contributor
Posts: 3

Re: Proc Glimmix Odds Ratio

Hi proc_fox,

I am having exactly the same issue in a similar model. Here's what it says in the users guide under GLIMMIX proc

OR<(odds-ratio-options)>

requests estimates of odds ratios and their confidence limits, provided the link function is the logit, cumulative logit, or generalized logit. Odds ratios are produced for the following:

  • classification main effects, if they appear in the MODEL statement
  • continuous variables in the MODEL statement, unless they appear in an interaction with a classification effect
  • continuous variables in the MODEL statement at fixed levels of a classification effect, if the MODEL statement contains an interaction of the two
  • continuous variables in the MODEL statement, if they interact with other continuous variables

I did some tests to see if I could figure out what was going on with my model. When I include only the types of variables described above, the exp(estimate) does in fact equal the odds ratio displayed in the OR table. However, as soon as I enter additional variables into my model (for which GLIMMIX can't calculate OR's) , the exp(estimate) and odds ratios no longer match.

Is it possible that the odds ratio table calculates and displays OR's as if only variables it can use (e.g. those listed above) are in the model, rather than displaying correctly adjusted OR's calculated for the complete model (including the interaction terms etc that it can't calculate OR's for.)? If so, the output would be extremely misleading.

Anyone else out there that can shed light on this issue? Which results should we trust - the calculated OR's or the estimated parameters?

Thanks.

Ask a Question
Discussion stats
  • 1 reply
  • 167 views
  • 0 likes
  • 2 in conversation