Table command exiting with error when using user-created program

[Chirag Yadav]

I am trying to make a table of normalized differences (based on Imbens and Rubin (2015)) between two definitions of income groups in my data. Here is an example dataset:

* Example generated by -dataex-. For more info, type help dataex
clear
input float(inc_status_1 inc_status_2 weight) byte raedyrs
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1  7696 10
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1 12696  9
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1  7361 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1     0 10
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  6930 12
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1  3364 14
1 1 10986 14
1 1 10986 14
1 1 10986 14
1 1 10986 14
1 1 10986 14
1 1 10986 14
1 1 10986 14
1 1 10986 14
1 1 10986 14
1 1 10986 14
end
label values inc_status_1 inc_status_1
label values inc_status_2 inc_status_1
label def inc_status_1 1 "Yes", modify
label values raedyrs EDYRS

This is my code:

ssc install _gwtmean, replace

cap program drop normdiff

program define normdiff, rclass
syntax varlist(min=1 max=1)
cap drop mean1
cap drop mean1_1
cap drop m_mean1
cap drop mean2
cap drop mean2_1
cap drop m_mean2
cap drop std1
cap drop std2
cap drop std1sq
cap drop std2sq
cap drop ndiff
cap drop std
cap drop m_std1
cap drop m_std2

qui egen double  mean1 = wtmean(`1'), weight(weight) by(inc_status_1)
qui gen mean1_1 = mean1 if inc_status_1==1
qui egen m_mean1 = max(mean1_1)

qui egen double  mean2 = wtmean(`1'), weight(weight) by(inc_status_2)
qui gen mean2_1 = mean2 if inc_status_2==1
qui egen m_mean2 = max(mean2_1)

qui egen std1=sd(`1') if inc_status_1==1
qui gen std1sq=std1^2
qui egen m_std1=max(std1sq)
qui egen std2=sd(`1') if inc_status_2==1
qui gen std2sq=std2^2
qui egen m_std2=max(std2sq)
qui gen std = sqrt((m_std1+m_std2)/2)

qui gen ndiff=(m_mean1-m_mean2)/std
format ndiff %20.10f
qui sum ndiff
local normdiff = r(mean)
return local normdiff `normdiff'
dis "The normalized difference is" %20.10f ndiff
end

normdiff raedyrs

table (command) (result), ///
        command(Diff = ( r(normdiff)) : ///
                normdiff raedyrs)

I am using Stata 17. The program works fine but I get an error in the table command saying "command normdiff exited with an error when run using the entire sample". I am unable to figure out what is causing the error.

[Chirag Yadav]

If anyone has any suggestions about tabulating stored macros from the program into a table, that would also help. I want to run "normdiff" on several variables and tabulate it together.

[Leonardo Guizzetti]

Evidently it is because -table- expects the command to support if and in in the -syntax- statement of -normdiff-. This is a good idea anyway to learn to use those qualifiers in your custom program. You can read more in -help syntax- on how to specify and use if and in.

[Chirag Yadav]

Thanks, Leonardo! I managed to get a table using collect instead. This is my code:

collect clear
quietly: collect , tag(Col[1]): normdiff raedyrs
collect layout (cmdset) (Col)  
collect style cell Col[r(normdiff)], nformat(%8.2f)
collect preview

I almost have what I want. But I am doing something wrong with the format for the decimal places. Any suggestions on fixing this code?

[Leonardo Guizzetti]

I'm not too familiar with the collect system, but here's what I did. First, you have to return numeric values as scalars. Return them as local macros means that -collect- treats them as strings, and as such, numeric formatting is irrelevant.

Change your return line in your program to

return scalar normdiff = `normdiff'

Then this code produces your table, though I'm sure there's probably a better way to do this part.

collect clear
collect get r(normdiff) : normdiff raedyrs
collect layout (cmdset) (result)
collect style cell result[normdiff], nformat(%8.2f)
collect preview

[Chirag Yadav]

That worked perfectly. Thank you!