Re: Sampling Sub-Populations [message #12531 is a reply to message #12527] |
Thu, 08 June 2017 10:12 ![Go to previous message Go to previous message](/theme/default/images/up.png) |
Bridgette-DHS
Messages: 3139 Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS Stata Specialist, Tom Pullum:
The DHS samples are stratified, usually by all combinations of region and urban/rural residence. That means the samples are intended to be large enough to provide relatively stable estimates for each stratum. Small strata tend to be over-sampled and then are weighted down; large strata tend to be under-sampled and then are weighted up.
At the national level, estimates by five-year age groups should be ok but this depends on the outcome. So far as sex (I would say gender is something else) goes, the emphasis is on women. Males are in the household survey and there is usually a survey of men (often smaller than the survey of women).
If you use weights, the samples are always representative for any subpopulation, but the smaller the sub-population, the wider the confidence interval. If there is some specific risk group you are interested in, you will need to look at specific surveys to see how stable the estimates are.
|
|
|