Tuesday, April 05, 2005

Flaky Statistics

This is why I don't pay attention to polls and most "reseach" you get in the popular press / interweb. It's crap because no one knows / cares about statistics enough to do any due dilligence on their methodology. Pew Internet Research put out a study indicating that 6 million people (out of all people have iPods/MP3 players) have listented to podcasts.This was widely reported in the press but if you read further, they asked 200 people and 60 of them said they had listened to some kind of internet broadcast (which is not podcasting). So the conclusion they drew from their own data is wrong in the first place, then you have to consider what the errors on extrapolating from 60 -> 6 million is based on the standard deviation of the original data. And then there is statistical bias in the way they collect the results. For example: How did they reach the people? Telephone? You can't do surveys on cell phones so you have to correct for that bias of people who don't have land lines or aren't home in the evenings (which you really can't do) and include it in the error for the final result - which you'll notice they didn't give. This is all a lot of effort and if anyone really was going to do this for what is a pretty stupid survey in the first place, I'm sure they'd conclude it wasn't worth the trouble.

0 Comments:

Post a Comment

<< Home