Tim Churches <tchur@OPTUSHOME.COM.AU> wrote [originally]:
> I am having some difficulties reproducing the results calculated by
PROC
> MEANS for a weighted std error of the mean. I am using the formula
> provided in the SAS documentation (see
> http://jefflab.queensu.ca/stat/sas/sasman/sashtml/proc/zormulas.htm )
> but it is not clear (to me at least) how teh std deviation is being
> calculated in the weighted case.
Hi, Tim! Long time no see!
I read your second post, so I see you have this solved. But I wanted
to toss in my $0.02 just to muddy the waters. Please bear in mind that
the WEIGHT in PROC MEANS doesn't properly address the underlying
problems
if you have real sample data from a sample survey. If you have the
classical problem of a smaple from a conceptually infinite population of
independent and identicallydistributed observations with normal errors,
and the weights only represent a summarization for convenience, then
PROC MEANS is great. If you have sample survey data from a finite
population, where you can't make the traditional 'general linear model'
assumptions, then you ought to be using PROC SURVEYMEANS.. which has a
very different variance estimator.
David

David Cassell, CSC
Cassell.David@epa.gov
Senior computing specialist
mathematical statistician
