Closed benthestatistician closed 1 year ago
As of f63f97b, new section "Accommodating moderator variables via absorption" (sec:accomm-moder-vari
) of estfun_DA.tex spec states the essential step. At least for the design-based calculation; I have to think about this more as it bears on model-based.
Correction: we're not currently absorbing moderator main effects. And we no longer have plans to do so: on review, we'll be able to keep the implementation cleaner and more coherent by only absorbing intercepts prior to estimation. See vignettes/not-for-cran/estfun_DA at b2ed390 or later.
We absorb main effects of moderators. Error from the absorption operation should be propagated down to vcovDA.