Quantcast
Viewing all articles
Browse latest Browse all 310

Difficult interpretation with PROC Glimmix

Hello everyone, I have another interpretation question regarding PROC Glimmix, this is quite complicated, I am quite lost. I have managed to fit the right model but it's impossible for me to make interpretations.

The model I ran is generalized mixed model. The model had an ordinal response with 5 levels (0-5) and 3 treatment groups being compared, where 3 is the new treatment and 1 and 2 are controls. I had some random effects, that's why it's a mixed model.

In the code, I used the next line:
Code:

estimate 'Score=0 v Score=1, Control 1 v New treatment' intercept -1 1 0 0 0 treatment -1 0 1 /exp cl;
This is a kind of a contrast. I have got an estimate, I did exp(estimate) and got an odds ratio of 0.05.

OR=0.05 , and the comparison is of control 1 vs. new treatment, but also score=0 vs score=1....

just to give you the direction of thing, a lower score is better, and the new treatment has more "zero's" (score=0) than the other controls.

How to I make interpretation of the estimate statement output in English, so a normal human will understand it, or at least someone who knows a bit statistics ?

thanks...

My output:

Label=Score=0 vs Score=1, Control #1-New Treatment
Estimate=-2.9764
S.E=0.9793
DF=25
t=-3.04
P=0.0055
exp(estimate)=0.05098

Viewing all articles
Browse latest Browse all 310

Trending Articles