Average rating: | Rated 4 of 5. |
Level of importance: | Rated 4 of 5. |
Level of validity: | Rated 3 of 5. |
Level of completeness: | Rated 4 of 5. |
Level of comprehensibility: | Rated 4 of 5. |
Competing interests: | The Reviewer is a Theme Editor for the journal |
This statistical analysis of DHS data should be using survey statistics (survey logistic regression), not asymptotic-theory “normal” logistic regression. Specifically, the authors should do a subpopulation analysis of the Rajasthan State data to correct the weights for exclusion of some of the survey sample. This is not a self-weighted random sample, it is a multistate cluster sample (from which a single state is being included in this analysis). Failure to use appropriate survey weights / survey estimation techniques (even before getting into subpopulation estimation issues) can sometimes lead to completely different conclusions from the “same” dataset. See for example: https://pubmed.ncbi.nlm.nih.gov/19713856/
The authors used SPSS to do their analysis; an overview of how to do survey estimation (the right way for these logistic regressions to be done) in SPSS is available here: https://digitalcommons.wayne.edu/cgi/viewcontent.cgi?article=3253&context=jmasm
Also in Methods 2.1, they have a “c” bullet point listed but the text is empty.
on behalf of Dr Matthew Gribble
Theme Editor, UCL Open Environment