lmer and glmer - problems if grouping variable is a factor with <3 data points per group

[tombisho]

lmer and glmer conver their model into a standard glm model and run the standard glm checks

so depression ~ age + (1|child_id) gets converted into depression ~ age + child_id and the glm checks run.

This is OK if child_id is an integer (quite likely because it will be in Opal unless someone has created a level for each individual!?). If child_id is a factor with ~<3 observations, then the glm checks fail.

If child_id is an integer, why can it not be converted to an integer? There do not appear to be checks to stop this is in asIntegerDS?

Otherwise, we could change the serverside lmer and glmer functions to strip the grouping factors before running the glm checks. We already don't return per id coefficients so this should not cause concerns.

[tombisho]

Factors can be converted to integers to solve this - confirmed by Tim. But long term it might be nice to remove the factor from the glm checks so that users can supply a factor or an integer