I began my graduate career in 1977, when statistical analysis was quite the thing in historical studies. I took stats classes, learned SPSS, all that stuff. My dissertation was going to be a statistical analysis of several hundred guild petitions to the city council of Augsburg, Germany. And it was.
In the course of doing that analysis, though, I discovered two things of lasting importance to me as a historian. First was that the sample sizes were too small to ever support serious analysis. As a fan of David Herlihy, I was sorely disappointed.
Second, and more important, was that what the numbers could tell me wasn't really very interesting. The genuinely interesting parts were the unique entries, the particular cases of individual craftsmen, and how those particular cases shed light on the society at large. I've come to refine that to say this: when it comes to history, only the least significant things can be described by numbers.
I realize the statement is sweeping, and perhaps "only" is a bit much, but the fundamental has remained with me. Even where big data (the new phrase for statistical analysis) is reliable--which it rarely is for pre-modern subjects--it's rarely very interesting.
