The perfect dataset

Major event: I turned in my event history analysis paper today. Better late than never, right?

Event history, like much of the latest trends in education research, relies on large datasets. I'm not keen on using large datasets collected by other people. Why? Because it's kind of like online dating, which I also don't like.

So you start looking around at the various datasets. Some are easy to reject, because they're too old, too short, not college educated. A few look more promising. You start a little communication - look at some more pictures, chat about variables. Maybe it stops there, because you don't really like their variables; maybe they break it off by refusing to communicate with you. (Sorry, no access to my data for you!) But sometimes things still look good, so you agree to go out. You meet for coffee, maybe dinner, try a kiss, run a few regressions. That's when you realize you aren't compatible. It can't give you what you need.

Still, better to figure it out when you're only committed to a few dates. The trouble is, you never ever find one you want a long-term commitment with, because it's always somebody else's dream dataset. (They have an open relationship, though, because data never gets jealous.) It's a search that's doomed from the start.

The problem with this metaphor is that the logical conclusion that the only way to happiness lies via Weird Science.