Certain questions regarding the analysis of longitudinal data seem to reappear every time an analysis plan is written. These include; the relative merits of random effects and marginal models, modeling of time trends, choice of correlation structures, choice of likelihood, corrections for degrees of freedom, missing data and others. When the response variable is binary or ordinal, there are even more choices to be made. In this roundtable discussion participants are encouraged to share their experiences and ‘lessons learned’ from clinical trials.