Robustness testing

[Chris Strickland]

We need more robustness testing for LMM, LinearModel and GLM classes.

[Clair Alston-Knox]

Agreed. I will seek some large datasets from Graham today, that will help sort LMM.

[Chris Strickland]

Also, check things work with weird options, for instance, no random effects, no fixed effects, different priors etc. Even purposely making mistakes, to see if we are getting sensible error messages is useful. If we aren't then, raise an issue (assigned to me or yourself if you want to fix it) and attach code that produces the unusable error message.