Failing: to signal the users that there might be something wrong! (Suitable for the default case)
Omit: omit all the missing cases. Linear models are used internally, which cannot deal with missing cases, so cannot do something for partially missing cases.
Exclude: the same as "omit" when fitting but return data with missing cases.
Complete cases should be defined by intersection of conmat, behav and confounds.
It would be better if
NA
can be handled here. Or a simplena.rm
argument could be added.