Tuesday, March 02, 2004

A day's not complete without polling news

The AP Stylebook's editor, Norm Goldstein, responded to my question about rounding sampling error.

Why the rounding?
Our current guidance splits the difference, because the effect of rounding the sampling error is doubled when we use the sampling error to determine statistical significance of the difference between two results (as spelled out toward the end of the AP Stylebook "polls and surveys" entry).
I could be wrong, but I think that means they would round it to the nearest point like they do poll results but fear that the rounding would be too magnified because the margin of error is doubled.

According to Goldstein, AP's director of polling says AP may revisit the rounding of sampling error soon.
He says he is a little leery of rounding sampling error to the nearest whole number, but not entirely comfortable with rounding to the nearest half-point either. So we'll review this and probably revise the Stylebook guidance one way or the other.
Remember, pollster junkies (I think there are two of you now): You heard it here first.


Post a Comment

<< Home