We work hard to maintain high retention in the Dunedin Study because a major threat to the validity of longitudinal research is non-random attrition. That is, the people who drift away first, or are hard to locate, or hard to re-assess are typically not a random group of participants. Indeed, they tend to be people within whom multiple difficulties aggregate. Thus they must be retained to faithfully capture the full range of life exposures and possible outcomes that occur in the general population. Coming close to this ideal helps strengthen our causal inferences.
