Obtaining the individual, relative variance contributions of fixed and radnom effects of an LMM in R

[eimichae]

Hello,
I am running a LMM (using the lmer4 package) with the following structure: lmer(Response~F1+F2+F3+(1|R1)+(1|R1:T2)+(1|R2))
For my research question I would like to quantify how much of the overall variance is explained by each indvidual fixed and each random effect. In other words I would like to be able to say something like: xy% of the overall variance are explained by the fixed effect F1, yz% of the overall model variance are explained be the interaction of the random effects R1 and T2 etc. Is there an R-package that can use my lmer models and provide me with the individual variance proportions (or another measure of effect size) for each fixed and random effect? The fixed and random effect variances should be comparable
I looked up several posts on stackoverfolw:e.g. https://stats.stackexchange.com/que...f-explained-variance-in-a-mixed-effects-model
But in most cases the recommended packages can only proved overall variances across all fixed or random effects (e.g. the MuMIn package) or they provide either only random effect variances (e.g. the specr package) or only effect sizes for fixed effects (e.g. anova_stats)
Your help is much appreciated Mike

 

[hlsmith]

Hmm, I am not well versed in this area, but I was thinking MM just could tell you how much variability each level explained. I don't recall seeing a value for each variable, like in multiple linear regression. Though, I could be wrong.

@spunky and @Jake do you have any input on this?

 

[eimichae]

Hi hlsmith,
Thank you for your answer.
You are partially right, LMM tells you indeed how much of the variability each random effect (& residual) explains. But it regards all random effect (&residual) variability as 100%
I would be interested in an approach that tells me how much of each random and fixed effect explains in terms of variability when the variabilitiy of all random effects, fixed effects as well as the residual together is considered 100%.