Comments on: A Lott of Old Rosh

By: ArchPundit

ArchPundit — Thu, 01 Jan 1970 00:00:00 +0000

What is really great is the margin of error for the subsample of DGU—+/- 20%

This of course assumes perfect survey construction, sampling and execution, which as your colleagues laughed about, isn’t too damn likely.

By: Bernard Yomtov

Bernard Yomtov — Thu, 01 Jan 1970 00:00:00 +0000

I’m not sure I understand this weighting issue. Are you saying that a young white male in CA is given more weight than a young black male in VT because there are lots more of the former in the population than the latter?

Then doesn’t the weighting depend on an arbitrary selection of categories that could strongly influence the outcome. Suppose both guys are 20 yrs old. Then you could weight them equally if you only considered age, or age and sex.

Am I misuderstanding something?

By: John Isbell

John Isbell — Thu, 01 Jan 1970 00:00:00 +0000

Someone pointed out that the missing repondents are the ones who were shot dead in the attacks. Funny how that warps the survey.

By: Atrios

Atrios — Thu, 01 Jan 1970 00:00:00 +0000

I can see no reason why any researcher would use those kinds of regional-demographic weightings for a national survey. None. Now, perhaps he did do it, but god knows why…

By: Three-Toed Sloth

Three-Toed Sloth — Thu, 01 Jan 1970 00:00:00 +0000

Small Worlds and Morbid Amusements

My lovely and talented wife is a statistics graduate student, currently employed as a research assistant to Professor Susan Murphy, of our statistics department over in the next building. Susan’s speciality is theoretical statistics as applied to socia…