Should we rely on the Kenward–Roger approximation when using linear mixed models if the groups have different distributions? | Publicación