simulateResiduals fails on glmmTMB model with rank_check = "adjust"

[gordy2x]

There is now a (optional) argument in glmmTMB that drops terms that make the design matrix rank-deficient (see https://github.com/glmmTMB/glmmTMB/issues/522#issuecomment-1472885734), this is what lme4 does by default. I can run simulateResiduals on an lme4 object but not on a glmmTMB object with rank_check = "adjust".

library(glmmTMB)
library(lme4)
library(DHARMa)
set.seed(101)
dd <- data.frame(z = rnorm(100), 
                 x1 = 1:100, x2 = 2:101,
                 re = rep(1:20, each = 5))

mod_lme4 <- lmer(z ~ x1 + x2 + (1|re), data = dd)
simulateResiduals(mod_lme4, n = 5)


mod_TMB <- glmmTMB(z ~ x1 + x2 + (1|re), data = dd, 
                   control = glmmTMBControl(rank_check = "adjust"))
simulateResiduals(mod_TMB, n = 5)

[reiniervlinschoten]

The issue is not in DHARMa, but in the predict function of glmmTMB, see my comment in the other issue. I think @florianhartig can close this as the problem is not in this repo.

[florianhartig]

Thanks Reinier ... I'll leave it open for now to check later if this has been fixed in glmMTMB, but will likely not address this in DHARMa.

[bbolker]

Resolved now with remotes::install_github("glmmTMB/glmmTMB/glmmTMB@rankdef_NA")

[florianhartig]

Hi Ben, many thanks! I'll keep this open until the problem is resolved in CRAN!

[reiniervlinschoten]

Awesome, thank you very much for fixing this!