Choice of the primary analysis in longitudinal clinical trials