I'm using jmp for my analysis, and I have data from what is ultimately a mixed model. The problem I have is that JMP doesn't have the ability to run GLMM. You can, however, run a mixed model with a standard least squares personality.

My question is: what am I missing out on by not having the GLMM functionality? What is the difference between what I've done (SLSMM) and a GLMM?

Clarification: response variable is continuous (%growth)

  • $\begingroup$ Can you please clarify the nature of your response variable? $\endgroup$
    – usεr11852
    Commented Feb 24, 2017 at 19:40
  • $\begingroup$ @usεr11852 The response variable is continuous (%growth) $\endgroup$ Commented Feb 24, 2017 at 20:04
  • $\begingroup$ Is it bounded in $[0,1]$ or are there instances where you can have say $100\%+$ or negative growth? $\endgroup$
    – usεr11852
    Commented Feb 24, 2017 at 20:48
  • $\begingroup$ @usεr11852 I have it in percentages and it ranges from about -26% to 80% $\endgroup$ Commented Feb 24, 2017 at 20:51
  • $\begingroup$ Hmm... OK. Yes, you are good with treating as a Gaussian (ie. through as LMM). In general if you had relatively small changes (say up to 25%) you could squint and say that that percentage chance can be approximated by change in natural logarithms. (You do not have small changes in the sample you describe.) See this threa too, I think it is very relevant. $\endgroup$
    – usεr11852
    Commented Feb 24, 2017 at 21:03


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.