Bivariate tests and median in a complex survey sub-sample

[Lara Guedes]

Hi folks.

I've been handling a complex survey dataset for a while now, and I can't find a way to compute median, SD, nor mann-whintey/qui/correlations tests for my weighted sub-sample. Any code suggestions?

[David Radwin]

The first part of your question is a FAQ: https://www.stata.com/support/faqs/s...r-survey-data/

Also see https://www.statalist.org/forums/for...35#post1358235

[David Radwin]

Also see https://journals.sagepub.com/doi/10....867X1201200202 and https://www.stata-journal.com/articl...article=st0198 .

[Lara Guedes]

Thanks again David.

Concerning SD, I managed to solve the problem.
Concerning percentiles, I'm trying to install epctile.pkg, as we were talking in the other post.
Concerning bivariate tests, thank you so much for the articles, specially the first one! It was really helpful to systematize ideas. Now, how can I compute the ranksum (or chis) in svy? Is it possible?

[David Radwin]

I don't know of any way to use ranksum with complex samples (svy setting). If by "chis" you mean "perform a chi-square test," that can be done with svy: tab, such as this toy example:

. webuse nhanes2

. svy: tab sex race
(running tabulate on estimation sample)

Number of strata = 31                            Number of obs   =      10,351
Number of PSUs   = 62                            Population size = 117,157,513
                                                 Design df       =          31

--------------------------------------
          |            Race           
      Sex | White  Black  Other  Total
----------+---------------------------
     Male | .4225  .0435  .0133  .4794
   Female | .4566   .052   .012  .5206
          | 
    Total | .8792  .0955  .0253      1
--------------------------------------
Key: Cell proportion

  Pearson:
    Uncorrected   chi2(2)         =    4.5159
    Design-based  F(1.93, 59.72)  =    1.2442     P = 0.2946