I have a mixed effect model, with two random effect variables. I wanted to rank the relative importance of the variables. The relimpo package doesn't work for mixed effect model. I am interested in the fixed effect variables anyway so will it be okay if I only take the fixed variables and use relimp? Or use weighted Akaike for synthetic models with alternatively missing the variables?
which one is more acceptable?
If I understood your question correct you want to provide some statement about the importance of your two random effect variables? You could provide the information of the explained variance of your two group variables.
It depends how you are going to define "relative importance"; there are a few ways you could do it. For example, you could calculate standardized coefficients (an easy way is to just z-score your independent variables and your dependent variables), and whichever predictor has the biggest standardized coefficient is the one that has the biggest numerical effect on the outcome.
Sofanit, You might try something like lasso or gradient boosting. See the attached. Hope this will help. Best wishes, David
A mixed effects model is a combination of fixed effects and random effects factors. Fixed effects factor levels are chosen specifically and not randomly. Statistical inference on them apply to the chosen levels specifically. Random effects factors are on the other hand chosen as random samples of levels. Inference does not apply to the chosen levels but to their underlying populations. Comparison may still be done by the use of the moments like the means: for fixed effects factors to the particular factor levels chosen and to the random effects to the population of levels, and not the specific levels.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics