Recently I have tried to come to grips with the Complex Samples module, and even if it there are relatively few steps involved, I can't say my usage so far feels safe. And all attempts to set up simple and small examples which should be easier to interpret than the real dataset have also become fairly complex, so I am still not convinced it works as I hope for.
The goal is to make comparisons between counties for different years. The data stems from a series of surveys which were all stratified on county, sex and age group. The sampling weights have been calculated with one single standard population, a procedure which as far as I can understand is termed post-stratification. I found an IBM support page from 2010 (
http://www-01.ibm.com/support/docview.wss?uid=swg21475022) saying that this was not possible. However, when I set up the CSPLAN with the Analysis Preparation Wizard, there is a step where I enter the population sizes. Comparisons made by not using the CSPLAN/CSTABULATE procedures, but rather WEIGHT and FREQUENCIES seem to give exactly the same results when it comes to estimated sample sizes. And the fine manual is not that detailed enough to make me see if I am doing something wrong.
Guess it had been good to have a firmer grip on sampling theory, but even so, the manual does not seem that helpful.
There are a number of questions here, but the first seems to be if post-stratification is possible at all with the latest version of the Complex Samples module?
Robert
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Robert Lundqvist
Locked
