Possible bug in .iter_remle function

[colinvwood]

I've run into an issue where I get the following error:

contrasts can be applied only to factors with 2 or more levels

from the lmerTest::lmer function, specifically the call here.

I called ancombc2 with the both a fixed effects formula and a random effects formula, specifically: body.site + (1 | subject). The body.site factor has two levels, and the subject factor has four levels. What happened is that the feature chosen by the expression y_crt = y[1, ] (see the link above) happened to only represent one level of the body.site factor, and thus the error above.

It seems like the first row of the feature table/OTU table is chosen as a matter of convenience, but by chance this feature may not represent multiple levels of the variables in the model. Choosing a different feature by changing the feature selection to y_crt = y[2, ] resolved the issue, and ancombc2 ran successfully. Note also that I did not encounter the issue with the same input data but no random effects formula.

Perhaps there is a way to make the call to lmerTest::lmer robust to such edge cases.

Possibly the cause of #271.