ABSTRACT
A longitudinal study collects measurements from the same set of individuals
repeatedly over time. This could be in a randomized clinical trial with intense
follow up for a short duration or in a panel survey with long term follow up
and a year or more gap between measurements. Analysis becomes complicated
when not all subjects participate at every time point, either by design or
by choice. Ignoring subjects who dropped out may bias some results. For
example, if the goal is to estimate the mean difference at the end of the study
and project it to a population, then subjects who are alive and lost to follow-
up need to be included. There may be some analysis where the population of
completers may be of interest. If the completers, however, are highly selected
then the analysis restricted to them may not be meaningful. Clearly, people
with missing covariate data cannot be ignored.
