glmSLMADS2.o does not have disclosure traps in it

[tombisho]

While working on SLMA of mixed models I have noticed that the glmSLMADS2.o function does not have the built in disclosure traps.
The glmDS2.o function has disclosure traps for things like over saturated models or categories with too few individuals.

Is this by design or does it need fixing? It would be useful to know if it is by design because then I won't have to do it for mixed models!

[pb51]

Hi Tom

Aaah!!! This is not by design - it is simply that I forgot to put in the traps which need to be the same as for glmDS2.o (or maybe it sounds better as a sort of nihilistic minimalist design on my part :o) ).

@demetris could you work with Tom to add these same traps from glmDS2.o into glmSLMADS2.o and resubmit the revised version to github. The required traps are for oversaturation, and then disclosive elements in the y vector, X matrix, or W vector. This is not entirely trivial, you will have to work out how to obtain the y vector and X matrix in testable form based on the code we used in glmDS2.o and the code in glmSLMADS2,o. If this proves very difficult, please feel free to leave till I return from holiday.

Cheers

Paul

[tombisho]

Hi Paul,

I've not heard of the nihilistic coding style before, but perhaps it will catch on ;)

For now I will continue with the lmerSLMA functions as I don't want to lose momentum on that. It is likely that the code I write for that could be used for glmSLMA because I think broadly the disclosure traps are the same (except when we are considering person id as the clustering factor).

Tom