The DHS Program User Forum: Weighting data » Sampling weights, cluster, strata in AIDS recode (AR)

Home » Data » Weighting data » Sampling weights, cluster, strata in AIDS recode (AR)

Show: Today's Messages :: Show Polls :: Message Navigator

Sampling weights, cluster, strata in AIDS recode (AR) [message #9025]

Wed, 27 January 2016 18:28

jgodwin
Messages: 5
Registered: January 2016
Location: Seattle

Member

I am attempting to fit some models and account for sampling weights usings the AIDS recode (AR).

I know the hiv05 variable represents individual weights and hivclust represents the cluster, but are there strata to be accounted for in the AIDS recode?

Specifically, I am working with KDHS2003 and KDHS2008 AR.

Thanks!

Jessica

Report message to a moderator

Re: Sampling weights, cluster, strata in AIDS recode (AR) [message #9033 is a reply to message #9025]

Fri, 29 January 2016 10:19

Bridgette-DHS
Messages: 3214
Registered: February 2013

Senior Member

Following is a response from Senior DHS Stata Specialist, Tom Pullum:

Yes, analyses of the HIV data should include the svy adjustment for strata, but the strata are not given in the AR file. You have to merge with the PR, IR, or MR file and get the strata from the file you merged with.

Report message to a moderator

Re: Sampling weights, cluster, strata in AIDS recode (AR) [message #9051 is a reply to message #9033]

Tue, 02 February 2016 12:44

jgodwin
Messages: 5
Registered: January 2016
Location: Seattle

Member

Just to be clear, after merging with another data set, by survey design object should use hivclust as cluster, strata from a merged data set as the strata and hiv05 as weights?

Jessica

Report message to a moderator

Re: Sampling weights, cluster, strata in AIDS recode (AR) [message #9065 is a reply to message #9051]

Thu, 04 February 2016 09:36

Bridgette-DHS
Messages: 3214
Registered: February 2013

Senior Member

Following is a response from Senior DHS Stata Specialist, Tom Pullum:

Correct. However, hivclust will be the same as v001 or hv001 (or v021 or hv021). They will be matched in the merge. You can think of the cluster id as being given by the AR file or the file you have merged with--it makes no difference because they are the same. The weight comes from the AR file (hiv05) and the stratum id comes from the file you have merged with.

Sometimes we want to look at the HIV status of couples, where both the man and the woman have been tested. In that situation you have hiv05 for the man and hiv05 for the woman. In that situation we use hiv05 for the man as the couple weight. The reason is that refusal rates are higher for men than for women. The weights include an adjustment for nonresponse. Getting the couple depends more on getting the man than on getting the woman, so it's more important to include the man's adjustment than the woman's adjustment. Of course you could define a couple weight differently, as you choose, but to match DHS results for couples you would use the man's value of hiv05 as the couple weight.

Report message to a moderator

Re: Sampling weights, cluster, strata in AIDS recode (AR) [message #9068 is a reply to message #9065]

Thu, 04 February 2016 14:27

jgodwin
Messages: 5
Registered: January 2016
Location: Seattle

Member

Thank you!

Jessica

Report message to a moderator

Re: Sampling weights, cluster, strata in AIDS recode (AR) [message #9607 is a reply to message #9065]

Wed, 20 April 2016 21:10

jgodwin
Messages: 5
Registered: January 2016
Location: Seattle

Member

Is there any documentation on the weighting of the AR?

I am trying to use the AR to estimate prevalence as a continuous, smooth surface while properly adjusting for the sampling scheme. I am struggling to find any documentation at all on the AIDS recode.

Thank you.

Jessica

Report message to a moderator

Re: Sampling weights, cluster, strata in AIDS recode (AR) [message #9613 is a reply to message #9607]

Fri, 22 April 2016 14:48

Bridgette-DHS
Messages: 3214
Registered: February 2013

Senior Member

Following is a response from Senior DHS Stata Specialist, Tom Pullum:

You can only analyze the AR file if you merge it with the PR, IR, or MR file. I have provided some examples of how to this in other forum posts.

Your goal of producing smooth surfaces of HIV prevalence, what we refer to as interpolated surfaces in some recent Spatial Analysis Reports, is too ambitious. The data are too sparse. You are welcome to try but we cannot help with that.

Report message to a moderator

Previous Topic:	Denormalizing data
Next Topic:	Regarding de-normalizing and weighting procedures in Stata

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Thu Dec 26 21:40:45 Coordinated Universal Time 2024