`check_predictions` not working for `nlme::lme()`

[mattansb]

library(nlme)

fm2 <- lme(Reaction ~ Days,
           random = ~ Days | Subject,
           data = lme4::sleepstudy)

performance::check_model(fm2)
#> Could not compute standard errors from random effects for diagnostic
#>   plot.
#> Homogeneity of variance could not be computed. Cannot extract residual
#>   variance from objects of class 'lme'.
#> Converting missing values (`NA`) into regular values currently not
#>   possible for variables of class `NULL`.
#> Error in d[, !(names(data) %in% all.varying), drop = FALSE]: incorrect number of dimensions

^{Created on 2022-10-23 by the reprex package (v2.0.1)}

[strengejacke]

That's because simulate() works, but produces odd results for check_predictions():

library(nlme)
library(performance)

fm2 <- lme(Reaction ~ Days,
  random = ~ Days | Subject,
  data = lme4::sleepstudy
)

check_predictions(fm2)
#> Error in min(replicated): invalid 'type' (list) of argument
simulate(fm2, nsim = 10)
#> $null
#> $null$ML
#>    info    logLik
#> 1     0 -887.5477
#> 2     0 -877.6983
#> 3     0 -886.0273
#> 4     0 -876.2553
#> 5     0 -871.8994
#> 6     0 -873.3890
#> 7     0 -864.8214
#> 8     0 -879.5240
#> 9     0 -880.4150
#> 10    0 -890.2539
#> 
#> $null$REML
#>    info    logLik
#> 1     0 -883.1471
#> 2     0 -874.0700
#> 3     0 -881.6479
#> 4     0 -872.4690
#> 5     0 -867.8706
#> 6     0 -869.5454
#> 7     0 -861.0608
#> 8     0 -875.2973
#> 9     0 -876.3774
#> 10    0 -885.8936
#> 
#> 
#> attr(,"class")
#> [1] "simulate.lme"

Should we just skip the pp-check for lme?

[mattansb]

Solution: https://stats.stackexchange.com/a/28627