Here is a question I received from a colleague:
I’m helping w/ a study where they want to convert the R2s to the liability scale, per Lee et al. We can do this for pseudo-R2s like Nagelkerke’s, but we’re unsure how to do this/is this possible with the marginal R2 from a mixed-effect model, per Nakagawa et al. Have you done this, or have any advice on this topic? Wondering if we can use the same transformation as for Nagelkerke’s R2, but that seems too easy…
I don’t know the answer. Any thoughts? Here was my response:
Short answer is that I do not know. I don’t think that converting the marginal R2 from mixed-effects models, ala Nakagawa, to the liability scale is straightforward. The marginal R2 reflects the variance explained solely by fixed effects, excluding random effects and residual variance, which are integral to the total variance framework assumed in liability scale transformations. I’m at the workshop this week so can try to ask people here if they have any ideas…
Any thoughts??