I am assessing cross-sectional repeated measures data using a mixture of linear mixed models (LMM) and Generalized Linear Mixed Models (GLMM). I see in various places that GLMM is used primarily for prediction of outcome variable values. My research is primarily to infer relationships between a set of independent variables and an outcome variable (inference) rather than prediction of values. Is GLMM still appropriate to use with continuous/approximately-continuous outcomes when the goal is inference over prediction?
3
-
1$\begingroup$ Please edit the question to provide more context and to provide links to the references ("various places") that suggest a difference between LMM and GLMM in this regard. It also would help to say more about the substance of your study; sometimes there are ways to analyze all the data together that might be preferable to the separate models you seem to be using. $\endgroup$– EdMCommented Aug 10 at 21:11
-
1$\begingroup$ Is "yes" a detailed enough answer? In my experience, GLMM is used primarily for inference. $\endgroup$– wzbillingsCommented Aug 10 at 22:21
-
1$\begingroup$ Seconding @EdM, I am always interested in seeing the links when people read stuff "in various places". I have personally thought about GLMMs as mainly an inferential tool. Where are people saying that it's used primarily for prediction ... ??? $\endgroup$– Ben BolkerCommented Aug 13 at 13:58
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Yes is good enough. I am using GLMM with robust estimation to compare with LMM results using automated syntax, both with an eye to inferencing the impact of a series of predictors on a series of outcomes. Thank you!