"I wish I had done this!"
Mood: geeky
Posted on 2008-11-26 10:16:00
Tags: essay math programming links
Words: 387

"I wish I had done this!" is my highest praise for a website. The last time I can remember using it was for wowjutsu, which tracks raid progression in WoW by looking up what gear people are wearing in the Armory and matching that with where that gear came from. Simple idea, useful, interesting, but the technology behind it is something I totally could have done.

My newest "I wish I had done this!" is StateStats. You enter a search term, it finds which states in the US search for that term more per capita, then gives you a nice heat map of the US. But then it correlates that with a host of other state rankings: obesity, income, high school graduation rates, voted for bush, percent youth, etc., etc., etc. So you can see that searches for "prius" are correlated with income and negatively correlated with energy consumption. Or searches for "gay" are correlated with density (i.e. more urban states) and negatively correlated with voted for bush. Or searches for "lsu" are highly correlated with, well, being Louisiana. Or searches for "coke" are highly correlated with obesity, while searches for "soda" are highly negatively correlated with obesity. (huh?) Or searches for "tea" are correlated with income and negatively correlated with voted for bush.

Anyway, it's a ton of fun to play with, and the example queries ("garth brooks" is highly correlated with voted for bush!) are interesting, but it's even more fun to think of a common search term and see what pops up.

The correlation metric it's using is just based on rank and not intensity - i.e. it's just the order of the 50 states that matter, not how much the first place one is bigger than the second place one. This probably leads to some false positives when the numbers are very close together. Also, I'd imagine there's a natural inaccuracy determining which state a particular query is coming from, and since you're looking at things only on a state by state level (as opposed to county by county or something) it's not as precise as it theoretically could be.

And don't forget correlation is not causation - searching for "hockey" does not make it colder outside, or make you richer.

I award StateStats the official "I wish I had done this!" seal of approval.

2 comments

Comment from othersider:
2008-11-28T01:17:55+00:00
I agree. I'd also like to bestow my personal "I wish I had done this!" to Where's Lunch?.

Comment from gregstoll:
2008-12-01T11:38:55+00:00
Aw, thanks :-) I think the next things I work on are filtering based on tags and general performance improvements...

This backup was done by LJBackup.