Casing it

So today is the next installment of our comparative politics seminar. This time we're discussing the issue of case selection and the potential pitfalls one can fall into. Having down the recommended reading (Geddes, van Evera et al), I'm staggered that there is so much science on the subject and the problems seem so great that it's a wonder comparative studies work ever gets done.

For example, van Evera recommends looking for extreme types of cases, since their accentuation makes it clearer to see the impact that one variable may (or may not) have on another. But Geddes notes something called 'regression from the mean', which if I understand it correctly, means that taking something extreme could indicate abnormality - and is therefore not an accurate reflection of reality.

Oh, what to do?