similarity matrix 2
A promise is a promise, even if no one particularly cares. Here's that DHTML similarity matrix I promised. I haven't done much cross-browser testing, but in Firefox at least you should now be able to see which state delegation are the most homogenous* and which are the most loyal to their party. The results won't shock you (Californians are better Democrats than Kansans!?!!), but it's nice to see that my math seems to be correct.
Now that I'm home (and Southby preparations are over with) I can devote some more time to this Sunlight Foundation contest. I've got a few ideas and a bunch of very good leads courtesy of son1. The deadline isn't until mid-April — I think I should definitely be able to come up with something cool by then.
* Yes, I know that I misspelled "homogeneous". I just feel very strongly that it ought to be spelled "homogenous". Milk isn't homogenie-ized, is it? Yeah, that's what I thought. WHAT!
UPDATE: Looks like the DHTML is a little finicky. Given that this isn't the final form that this data will be visualized in, I don't think I'm going to bother fixing it. It still works on the permalinked page though, so if you want to see it in action follow that link.
Comments
That's pretty great. It's a Processing thing, right?
I should have offered in the other comments, btw -- I've got some simple Java implementations of a couple clustering algorithms lying around (for stuff that I needed to do for work). I don't know how useful they'd be, but if there's something specific you want to do I'd be happy to pass along some source code (or maybe even a compiled tool or two).
Nah, this isn't Processing. I thought about doing it in that, but I was in the mood to play with Javascript. So this is just JS and HTML.
I'd love to see whatever you've got -- Processing can use genuine Java source code, so who knows, I may be able to use it directly. But I'm even more interested in seeing the algorithms and trying to trace through how everything works. Thanks very much for the generous offer.
It'd be nothing. I've got the code just sitting around, it'd be a shame if I was the only one to get any use out of it. I'll take a couple of days to look around and clean it up, and then I'll post it somewhere where you could grab it.
I also have an interest (both professional and personal) in graphical models, and I've been kinda mulling over how you might use them to build/learn an appropriate model for this kind of data. So if any of that looked interesting to you as well, you should let me know -- I'd be happy to (again) point you in the right direction, and maybe even pass along the tools you'd need along the way.