| District-level aggregates [message #12332] |
Tue, 02 May 2017 13:42  |
Christian Bommer
Messages: 13
Registered: June 2015
|
Member |
|
|
Dear DHS team,
I realize that there have been questions before on the representative of DHS data. The position of DHS on this (please correct me if I'm wrong) is that surveys were (in almost all cases) designed to be representative at the national, regional and urban/rural level. However, I was wondering why this is the case. Specifically, what I want to do is to use DHS data in order to estimate the population share of different ethnic/religious groups at the district level. I am currently unsure whether this would be really problematic to do. I can imagine two potential issues and would kindly like to ask you whether these are the only reasons why DHS advices against district-level analyses:
1) Power: There maybe a number of districts for which only very few clusters where selected such that the margin of error maybe substantial. I believe, however, that this is not necessarily an issue of representativeness but rather of power, i.e. this should be reflected in the CI (provided that a minimum number of clusters is available).
2) Relocation: To protect privacy, clusters were displaced randomly in a radius of 10 km maximum (please note that my question only refers to DHS surveys which were also geocoded). Clusters originally located close to district A may therefore end up in district B. However, if this is the only problem, I would argue that it is possible to simulate the potential bias and construct a measure of uncertainty (I probably only need to know the ethnicity shares with an margin of error of +/-5 percentage points).
Is there anything else that I am missing? I would very much appreciate your thoughts on this.
Kind regards,
Christian
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
