| SAS: applying female weight [message #2571] |
Tue, 15 July 2014 17:25  |
gnprieur
Messages: 1
Registered: July 2014
Location: Atlanta
|
Member |
|
|
Is it necessary to apply any other sample weight if I'm only looking at data for females?
Also after applying weight is it appropriate to do Chisqr or binomial tests?
Just went ovet this today with my instructor and was told not to run this on survey data collected as clusters.
|
|
|
|
| Re: SAS: applying female weight [message #2576 is a reply to message #2571] |
Wed, 16 July 2014 12:51  |
Bridgette-DHS
Messages: 3199
Registered: February 2013
|
Senior Member |
|
|
Please see the following response from Senior DHS Specialist, Tom Pullum:
The appropriate weight to use is v005 in the IR, BR, or KR files; hv005 in the PR or HR files; mv005 in the MR or CR files; hiv05 if you are using the HIV data linked to any other file.
The only exception is that for surveys of ever-married women, rather than all women (age 15-49 in both cases), you need to adjust v005 with an "all-woman factor" if you want to interpret the results as applying to all women, not just ever-married women. The easiest way is to look at the variable awfactt (the second "t" is for "total") and see if it is 100 for all women or it varies from one woman to another. If it varies, then this was an ever-married sample and you need to calculate the weight as v005*awfactt (or v005*awfactt/100; usually the factor of 100 won't have any effect because the weights will automatically normalize to a mean of 1). There are other versions of awfact, with a different last letter. You can look up which one to use in the Guide to DHS Statistics. There are only a few surveys now in which awfact is relevant.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
