Sunday, October 12, 2008

Let's go to the numbers!

Over at More Grumbine Science colleague Robert G. has an excellent discussion of precision in data, and how it can often be driven too far. This is an especailly prescient thing for us scientists types to pay attention to in an election season, especially sinmce we get bombarded more an dmore by "figures" and "numbers" that purport to tell us somethign about this or that candidate.

Excess precision, and manipulation of statistics, gets all the more interesting when you read Darrell Huff's How to Lie With Statistics. I owe Robert a huge Hat Tip for that one as well, though if you blow by his first paragraph, you'll probably miss the reference.

After reading the little tome, and surviving a graduate minor in statistics along the way to my M.S., I have to admit taht I'm having a lot of fun with statistics this campaign. I leave it to others to dig into the facts - I just don't have that kind of time. INstead, I'm focusing on looking hard at the analyses of the numbers, and making sure that when I hear inaccuracies bandied about, I correct them. I also strive to tell folks why a number i sincorrect, becuase I'm firmly convinced that too few Americans are sophisticated enough to figure out the methodological twists themselves. Of course, having said that, I'm sure I'll be accused of being a liberal elitist.

Take, for instance, the often repeated and analysed claim that Sen. Obama voted to raise taxes 94 times. By now, many outlets have noted that the vote tally actually consisted of 23 votes against tax cuts, 7 votes to raise taxes on wealthy individuals or corporations, and 17 other votes were on 7 bills, mostly to amend them in one way or another. Bottom line - yes tehre wer e94 votes, but no, they don't add up to 94 seperate actions to raise taxes.

How is this lying with statistics you ask? Well, for starters, we are never told what the universe of Sen. Obama's votes is. Lookign it up through Google, we find that since taking office, the Senator has had the chance to vote on 1299 actions befor ethe Senate. He has, interestingly, missed 314 of those. That percentage, interestingly, didn;t get very big until Q2 of 2007 - once the Presidentail campaign began in earnest. So even if we looked at the "94" votes as separate legislative actions (which they are not), they would only be 7.23% of Sen. Obama's record. Worse yet, fo the 94 votes in question, 24.5 % were against tax cuts, which means that 1/4 of the time Sen. Obama was being . . . . wait for it . . . . . a Democrat who believes in taxing people for the services they enjoy from the federal government. How sad that Republicans, supposedly the "Conservative" arm of our political system, still think that forcing people to pay fo rthe services they are provided makes you a "tax and spend" liberal.

My point in all this is that modern campaigns are amazing whizzing machines at chruning out numerical assaults on opponents. Sure, out of context, 94 votes to "raise taxes" wounds bad. But when you get into the numbers, and you do some further testing, they tend to break down. Sort of like Huff's story about the "average Yale Man" making $27,000. huff's point - which Ithink you should read in his book, is that without knowing whether "average" refers to the mean, the median, or the mode, you can't really tell what that $27,000 actually means. Sounds a lot like the numbers game the campaigns are playing, and getting away with.

No comments: