3

I'm estimating a multilevel ordinal logit with one interaction term using clmm. The coefficients I get are interesting and statistically significant and I now want to compute and plot predicted probabilities. In doing so, it may be relevant to consider that all my independent variables are categorical. How can I compute and plot predicted probabilities for models estimated using clmm?

I have looked online relatively extensively and can't find a way to compute and plot predicted probabilities for clmm models. The "effects" package does not seem to work with clmm either.

This is the code I'm using to estimate the model:

mm3c = clmm(var1 ~ var3 + var4 + var6 + var2*var5 + var7 + var6
            +  (1|var8), data=my.data)

My goal is to estimate and plot probabilities for the interaction of var2 and var5. Any help would be very appreciated--thanks for your consideration!

saladin1991
  • 142
  • 9
  • any chance we can have a [mcve]? – Ben Bolker Apr 24 '19 at 17:25
  • Hi Ben : Thanks a ton for your consideration! I did not associate my post with any example as my problem isn't based on data specific issues. It's rather a package specific "problem", so to say. As such, any self generated data that is similar to my variables --ordinal DV and categorical IVs--should do the job. "Effects" packages and "predict" function won't work simply because they aren't compatible with clmm. I was wondering what folks use to predict probabilities, compute odds ratios, or marginal effects when using the clmm package. Thanks again for your help! – saladin1991 Apr 25 '19 at 14:41

0 Answers0