|Date: ||Mon, 18 Jul 2011 13:20:43 +0000|
|Reply-To: ||"Fehd, Ronald J. (CDC/OCOO/ITSO)" <rjf2@CDC.GOV>|
|Sender: ||"SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>|
|From: ||"Fehd, Ronald J. (CDC/OCOO/ITSO)" <rjf2@CDC.GOV>|
|Subject: ||Re: PROC SURVEYFREQ SE versus SUDAAN|
|Content-Type: ||text/plain; charset="us-ascii"|
> From: Marc Mooney UMN
> Sent: Monday, July 18, 2011 1:11 AM
> To: SAS-L@listserv.uga.edu
> Subject: PROC SURVEYFREQ SE versus SUDAAN
> Hi all,
> I am doing an analysis on a large public dataset. Previous reports on
> this data set have used SUDAAN. In replicating some one-way
> frequency tables, I can exactly replicate estimates of percentage
> rates, BUT my standard error of percentages are much smaller
> (about 1/3) than those from SUDAAN.
I attended a pre-conf seminar several years ago, 2009,
comparing the SAS Survey* procs with Sudaan.
My limited (non-statistician) understanding of the difference is the way that
missing values are handled.
Using Replication Methods to Analyze Survey Data in SAS(R) Software
Anthony An and Donna Watts
there are a couple of google hits for this doc:
Ron Fehd Sudaan (sigh) maven
> As a for instance:
> proc surveyfreq data=temp;
> tables AVOIDPDX2 /alpha = .01;
> strata stratum;
> cluster psu;
> weight weight;
> I am assuming I am needing to specify something else, perhaps
> about variance estimation. I appreciate any suggestions.
> Thank you,