Wilcoxon rank sum test + Benjamini-Hochberg correction

[Lilian Los]

Dear Statalist Users,

I am currently working on analyzing missing data in my dataset through several approaches (e.g., including performing an mcartest that did not converge). As a result, I am currently working on analyzing the missing data by comparing the values in each variable (columns) between observed and missing data (reported as a dichotomous variable 0=observed, 1=missing) in all the other variables (rows) using the asymptotic Mann-Whitney U test (i.e., the ranksum command) followed by the Benjamini-hochberg correction method, a process described in this paper for example (please see relevant supplement here for what I really aim to do).

However, as I have around 40 variables that have missing data, I was wondering if there was a more efficient way to conduct my analyses?

Currently, I am planning on running the code below for each pair of variables, followed by correcting for the p-values manually on Excel (i.e., ranking the p-values in each column and then applying the correction formula). However, this seems inefficient with so many variables and p-values to go through, and as I am still new to STATA, I was wondering if there was a more efficient method?

ranksum cesd1, by (cesd1_miss)
ranksum cesd1, by (cesd2_miss)
ranksum cesd1, by (cesd3_miss)
ranksum cesd1, by (sleep1_miss)
// etc...

Again, here is an example table of what I am ultimately hoping to create:

	cesd1	cesd2	cesd3	sleep1	sleep2	dm1	dm2	etc...
cesd1_miss	corrected P-Value (CP)	CP	CP	CP	CP	CP	CP
cesd2_miss	CP	CP	CP	CP	CP	CP	CP
cesd3_miss	CP	CP	CP	CP	CP	CP	CP
sleep1_miss	CP	CP	CP	CP	CP	CP	CP
sleep2_miss	CP	CP	CP	CP	CP	CP	CP
dem1_miss	CP	CP	CP	CP	CP	CP	CP
dem2_miss	CP	CP	CP	CP	CP	CP	CP
etc....	CP	CP	CP	CP	CP	CP	CP

Thank you all!