I am doing a research in which i am trying to measure the importance of the doctor that is in charge of a patient in a medical decision. For that (and others reasons) i have used a GLMM using lmer4 that looks like that :
FitGlmm <- glmer(medical_decision ~ fixed_effect_1 + fixed_effect_2 + ... + (1|`dr`), family=binomial, data=data)
What i would like is to have a way to compare the "importance" of each effect (fixed AND random) in the medical decision.
From my (poor) understanding, pseudo-R2 (despite their limitations) could be of help to answer my question. After looking into partR2 and rptR packages, my issues, since i want to compare random effect vs fixed effects, is the following :
- Can i use ICC (or R) from rptR package to measure the proportion of variance of the random effect (dr) and compare it with R2m and partR2 of fixed effects ?
- Can i use this homemade formula
R2c - R2m(getting R2c and R2m from partR2 package) to get the proportion of variance of the random effect "dr" ? And then compare it with partR2 values of fixed effects from partR2 package ? - Should i use a completely different approach ?
Thanks in advance, Thomas
