Why you can’t trust statistics

Wired has an article on seven skills you didn’t learn in college. While some of them are silly skills (the art of the tweet? Seriously?), the writer did get the first skill exactly right: statistical literacy. As a holder of a math degree, this would be the first on my list as well.

This CNN article reeks of someone who never took statistics. The title draws you in: “Was Bush better president than Obama?” No matter what you think, you’re probably going to read it.

The article gives you a bunch of statistics comparing Obama to Bush and McCain and Biden to Palin. The really significant bit is the headline statistic: “By 47 to 45 percent, Americans say Obama is a better president than George W. Bush.”

That’s pretty close. Doesn’t that make you wonder how accurate these statistics are? Let’s see if the article says anything.

“The CNN/Opinion Research Corporation national poll was conducted October 5-7, with 1,008 adult Americans, including 938 registered voters and 504 likely voters, questioned by telephone. The survey’s overall sampling error is plus or minus percentage points, with a sampling error of plus or minus 4.5 percentage points for likely voters.”

That’s right. A 4.5 percentage point sampling error for likely voters. No sampling error is given for any other group in this survey. However, based on this I would put very little trust in what Americans say about who the better president was and make up my own mind.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.