Thursday, July 05, 2007

Bad Statistics

I make a living using statistics, it's my favorite science. I think there are two types of people in the world - those who have a fundamental grasp on statistics and... the other people. What gives statistics such a bad name is that they require so much common sense to interpret the data. The math is correct, the statistical theorems are all solid mathematically. But it's studies like this one that make people suspicious of statistics altogether. Well I'm not necessarily criticizing the study itself but the way the Associated Press is running with it.

WASHINGTON - Another stereotype — chatty gals and taciturn guys — bites the dust. Turns out, when you actually count the words, there isn't much difference between the sexes when it comes to talking.
Funny how since contemporary society (especially the left) ignorantly hate all stereotypes are using the principal science behind stereotyping - statistics - in order to disprove one!

The researchers placed microphones on 396 college students for periods ranging from two to 10 days, sampled their conversations and calculated how many words they used in the course of a day.

The score: Women, 16,215. Men, 15,669.

The difference: 546 words: "Not statistically significant," say the researchers.

So a joe schmoe reading this article may think the issue is settled when it's far from it. You can read further to get some more clarification but let me just interpret what this really means: it means that this particular survey did not produce enough evidence to say that college girls (in that particular area) spoke more than their male counterparts when they knew they were being recorded for survey purposes. That's what "not statistically different" means. They may very well be different and in fact, by all observable evidence they probably are, but we simply don't have enough evidence from this survey to conclude that they are with 95% confidence or more. So we may be 94% confident that they talk more but it's not confident enough to meet the .05 alpha which is the typical standard.

It may be true that they don't talk much more I don't know. From my own personal observances I would bet that it's not true, but all I'm saying is this survey doesn't disprove the stereotype.

No comments: