The letter in today's picture almost wrecked my presentation in Oxford last week - and reminded me how easy it is to get quantitative papers spectacularly wrong. What happened? While re-writing my talk two weeks ago (basically extending earlier work from Lucknow to the whole of Uttar Pradesh), I noticed some odd phenomena in my dataset. And freaked out. Oxford is not just any place to be invited to speak at. I had to know quite exactly what had gone wrong. After some fairly tense hours, I discovered it: the "Š", a letter which apparently crept into some electoral rolls of Eastern UP (and no, these are not written in latin script, but in devanagari). I have no idea why they were there in the first place, and there seemed to be no system - but as soon as my name-matching algorithm stumbled across them, it crashed. And left my dataset corrupted. Luckily, I was able to solve the problem (the final dataset arrived literally five minutes before my presentation), and could ditch the "I am truly sorry but my talk just dissolved in a data nightmare" embarassment. Close call!

Similar to earlier data trouble, I thus realized again how easy it is to spectacularly fail in quantitative research. You get one calculation wrong, a data row slips elsewhere - and your analysis is blown. It's much harder to fail equally grand in qualitative research. This is no argument in the quantitative-qualitative debate (which I find silly most of the time anyway). But if you deal with numbers as part of your research: be careful. Very careful. There might be a lurking "Š" around, waiting to destroy your fancy arguments at the most inconvenient hour...