re: #269 jaunte

2400 respondents is a ridiculously small sample.

It depends on how they were chosen. If they were selected completely at random from a pool of potential respondents who were uniformly distributed across the spectrum of possible responses, then the chances of obtaining a significantly different result from similar polling would only be about 2%.

The trick is in how the sampling was done. Was it truly random? Was the population it was drawn from truly representative of the population as a whole? Impossible to say, in this case, due to a lack of information.

Examples: if the pollster rejects certain answers - like Democrats responding “Nirhterism rock!” - then the polling isn’t random; it is biased. Most bias problems are far more subtle, but they certainly exist. In the second case, if the population surveyed consists of subscribers to the New York Times it is reasonable to conclude that this group is representative of the population as a whole, although it may give an accurate representation of what that particular sub-population thinks, assuming there wasn’t polling bias present as well.

Frankly, I’d like to see more data points before accepting these results. Some serious questions were raised yesterday regarding the provenance of this poll.