Dear RG Members.
I am using the Demographic & Health Surveys (DHS) data in a research about children's maltreatment. As you know, DHS data have a complex sampling design with strata and clusters.
I am trying to do regression models for explaining maltreatment by socio-economic variables.
DHS instructions explains that it is not necessary to adjusting for sampling design when doing regression models; opposite to some statistical books which suggest that (e.g. using svy: commands in Stata).
What is the right approach for modeling these regressions?
Thank you very much.
Yes you should take survey design into accourt in your analysis. In particular, an inapproriate specification of the random structure of your model may lead to incorrect conclusions (spurious statistical significance).
Andrés -
You note this: "DHS instructions explains that it is not necessary to adjusting for sampling design when doing regression models...."
My guess is that they have noted that it will not make a huge difference in this case. You could try both ways and see if that might be the meaning here.
If you want to explore this further, I think you might find one or more of the following books helpful, if I understand your question correctly:
Särndal, CE, Swensson, B. and Wretman, J. (1992), Model Assisted Survey Sampling, Springer-Verlang.
Brewer, KRW (2002), Combined survey sampling inference: Weighing Basu's elephants, Arnold: London and Oxford University Press.
Also one by Valliant, Dorfman, and Royall, circa 2000.
Cheers - Jim
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Applied Statistics
- Survey Design