svyset WBES

[Nancy Barakat]

Hi all
I am working on my master thesis tackling the impact of firm-level adoption of mobile money (a binary variable) on its energy intensity (a variable that ranges from 0 to 1). I get the data from World Bank Enterprise Survey, yet I am wondering if I should svyset my data before initiating with any modelling, that is to identify the weights and stratum variables. If so, please guide me through it.

Thanks in advance

[Daniel Schaefer]

Whether or not you should account for complex survey design is something you should discuss with your thesis committee. Just keep in mind the -svy- prefix will work for many, but not all commands. Check the documentation for details, and read the codebook for your survey to see how to -svyset-. Many will provide the exact Stata code you will need for svyset.

[Nancy Barakat]

Thank you for your reply. I checked for it and indeed I find the exact code, but as you have said svy doesnt work for "xtreg,fe and xtivreg". In this case, how can I run panel model with svy?

[Daniel Schaefer]

If the model doesn't support the svy prefix, you can't use that model and account for the design. Some of these modeling issues are open questions in Statistics. The real bleeding edge complex survey design stuff tends to be implemented in R first, but some of those packages haven't even gone through proper code review. I would avoid that kind of thing for a masters thesis.

Seems like you'll either have to drop the complex survey design or pick a different model. Check the summary statistics, bivariates and so on with the svy prefix and without. If the results are very similar, you may be able to make the case to your committee that you don't need to account for the complex survey design anyway.

[Nancy Barakat]

Thanks a lot for your help. Highly appreciated