pls I need guide on how to have as a single data extract of 30+ countries for many variables.
will kids be merged already with household persons and mothers to derive child nutrition status?
how do I ensure variables are standardized cause of pooled analysis?
can I also bysort for each country analysis in the data extract file?
do I also have to apply weight?
Pls I will appreciate every assistance regarding my above questions.
Thank you all.
It sounds like you’ll want to use the IPUMS DHS project. In this project, you can extract data at the child level. These children will be sorted and grouped with their households. Much of the work of IPUMS is to “harmonize” variables across time and space, which allows for the sort of “pooled” analysis you are describing here. In the extracted file, you’ll be able to perform all sorts of commands, such as bysort, that allows analysis at the country level. On this note, it is sometimes easier to have individual data extracts for each country, if you intend to perform analysis within countries.
Regarding sampling weights, if you intend to calculate statistics that are representative of some geographic area then you’ll need to apply sampling weights. A nice primer on sampling weights is detailed in this blog post.
Additional guidance on the use of IPUMS DHS is available here under the User Notes.
Jeff has already answered your basic question, namely, that you should be able to get the data you need using IPUMS-DHS as your data source. You select the countries, sample years, and variables you need, and the IPUMS extract system creates a single customized data file in the format you choose, with however many variables you specify.
IPUMS-DHS merges the household record to the woman’s record and also merges the mother’s record (and her household record) to the child and birth records, so you shouldn’t have to do any merging to get the information on mothers and household characteristics.
Some technical variables that help you determine which country and sample year an observation comes from are automatically added to your data file. These identifying technical variables include SAMPLE, COUNTRY, and YEAR. You can sort the data using your stats package with those variables.
You should definitely use weights. For the domestic violence variables, one woman (who had ever been married) was randomly selected for each household, so you need to use the special domestic violence weight when employing those variables (DVWEIGHT). If you are not using the domestic violence variables, you can use the standard person weight (PERWEIGHT).
Good luck with your research.