The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Sampling and Weighting Webinar June 2015 » Weighting for health outcomes
Weighting for health outcomes [message #5477] Thu, 28 May 2015 15:57 Go to next message
Gillian is currently offline  Gillian
Messages: 11
Registered: June 2014
Location: New York
Could you please talk about weighting for different types of outcomes? For example, a binary response like for diarrhea (0/1) seems pretty straight forward, but what about something like malnutrition where you want to use z-scores in categories, for example:

-4.00 to -3.00 Severe wasting
-2.99 to -2.00 Moderate wasting
-1.99 to -1.00 Marginal wasting
-0.99 to 2.00 Normal weight for height
2.01 to 6.00 High range weight for height

Thank you!
Re: Weighting for health outcomes [message #5480 is a reply to message #5477] Thu, 28 May 2015 17:55 Go to previous message
Messages: 292
Registered: March 2013
Senior Member

Do you want to estimate proportions? You could just use the "svy: mean" command after creating dummy variables for each group. But in general, the weights work exactly the same with continuous or binary variables (in fact, you are just turning one continuous variable into several binary variables with the setup you describe).

So generate a variable called "severe_wasting" which equals 1 if -4<HAZ<-3 and then compute the mean of that variable using the "svy" prefix and you'll get the weighted proportion of people in that group.

Note though that you will have some people who are in none of those categories (namely, people with HAZ between -6 and -4, which can be quite a few in some of these countries).
Previous Topic: weight in Probit, Logit models
Next Topic: Weights with HIV data
Goto Forum:

Current Time: Sun May 26 18:23:24 Coordinated Universal Time 2024