(no subject)

[lavendersparkle]

So, I've dragged myself into the faculty to work on my hated econometrics project. I fiddled with it a bit yesterday but didn't really get any results. I think that I need to do what professional econometricians do and just run shit loads of regressions until one of them comes out as vaguely statistically significant. Of course, this is a completely statistically invalid way of doing things as, 5% confidence intervals mean that if you ran a regression between two completely unrelated things there would only be a 1/20 chance of it being within the 5% confidence interval. So, if I spent all day here and ran 100 regression on unrelated things I'd expect about 5 to just happen to be statistically significant by fluke. If I were an econometrician I would then forget about all the regressions I ran that weren't significant and write a nice paper about how sun spots actually influence the exam results of second year undergraduate anthropologists.

Well, the course booklet did say we were supposed to write a paper like one that would be published in an academic journal so I suppose I better get on with data mining just like real econometricians do.