Prev: allow xcmd SAS eg 4.2 on windows 7 64-bit
Next: Macros-- Help
From: John on 30 Jun 2010 08:55
Hello All,

I'm analysing a dataset that is based on a national survey. The
contractor that prepared the dataset provided final weights and
replicate weights. The RepWeights were created using a JK2 estimator
in Wesvar.

I'm analysing the dataset using SAS 9.2. Proc Surveyfreq and Proc
Surveymean are being used with varmethod = JK. There are 935 records,
702 postitive final weights. There are 100 replicate weights: the
repweight option is being used in the SAS PROC. The contractor
provided point estimates for all varialbes and variance estimates on
some variables. The point estimates do not always match, and all
variance estimates do not match. The contractor suggested that it was
due to the differeneces in SAS using a JK1 varience estimator and
Wesvar was using a JK2 estimator.

The contractor is suggesting using the option JKCOEF = 0.999999 to
account for the difference between SAS and Wesvar. Does this seem
like a reasonable approach to account for the difference between the
JK1 and JK2 estimators? Will this account for the differences in the
point estimates and variance estimates? Isn't all the infomration
related to the procedure to create the Replicate weights contain the
information that would account for the difference between JK1 and JK2?

Thank you in advance for you assistance,
John
 | 
Pages: 1
Prev: allow xcmd SAS eg 4.2 on windows 7 64-bit
Next: Macros-- Help