[vharuck]

I've been repeating this in my work with statewide health data. The answer to, "Is there a difference in X between group Y and group Z?" is always "Yes." No data collection or significance test needed. If you want to know if being Y or Z affects X, you need a properly controlled experiment.

Studies using observational and non-comprehensive data collection have their place. They show reality. If people in one city have an incredibly high rate of dying from heart disease, that means we could reduce premature deaths there. And we'd do that by looking at the results of controlled studies into the causes of heart disease, then see which ones are prevalent and changeable.