Tuesday, July 22, 2008

The Margin for Error

If you run a poll, you will never receive 0% for any answer. There is a certain percentage of people who will answer randomly or will mis-read a question, or will answer counter to their belief based on some expectation they have about the poll. Or whatever.

In a recent poll about Obama's religion, I was unsurprised to find that some large percentage of people knows Obama is a Christian, some large percentage believe he's a Muslim, and some smaller fraction don't know. 1% said he was Jewish.

This is what margin of error is for.

4 comments:

shadowfax said...

I saw that too. I thought it was hysterical. I mean, seriously, Jewish? So if it was a standard-sized national poll, it probably had about 1000-1500 respondants, so there were about 15 people who thought he was Jewish?

Bizarre.

Matt Dick said...

1545, I think. 15 people erred in some way is more likely. That's what margin for error is. In fact, I took a whole class in my master's program that was about surveys and polls and the science behind them. I finally understand margin for error, selection error, biasing errors, etc. and how to interpret and calculate them.

Interesting class.

But it's likely that someone was both interested enough to take the poll and ignorant enough to figure Jewish was correct.

James Parsons said...

Did your class also cover margin OF error?

Matt Dick said...

Yes, Margin of Error was covered.