Reply #4: Okay, I'm an epidemiologist but have a master's in biostatistics

[Snow]

and teach introductory grad-level courses. Statistical techniques allow you to state whether the finding you've observed is a result of chance, with a probability assigned to that statement. Anytime you see an association, the name of the game is ruling out alternative explanations, and the explanation of "chance occurrence" is one of the first things you want to rule out. Therefore, it's unfortunate that a bit of statistical jargon is the term "significant". That doesn't mean "important"; it simply means the degree to which you can exclude chance as an explanation. A very low probability of chance as an explanation doesn't mean you'v explained the finding yet. So, yes, you need a statistician, and yes, you do need poli-sci, polling, or maybe even better, a geographer, since geographers study this kind of thing all the time.

When you go looking for a statistician, I'm betting you'll need someone who knows how to deal with messy data - ie, the data may be sparse in some spots, it may not follow a normal distribution, there may be dependence among different variables. Who ya gonna call? Well, someone who knows and wouldn't mind using bootstrap, resampling, and Monte Carlo analytic techniques. Maybe even (and here half the statisticians in the world will shudder in horror) someone who's a bayesian. Good luck!