Some commands appear to ignore [w=weights] about stata-gtools HOT 3 OPEN

@erciomunoz I've decided to leave this issue up in case others have a similar question about how gegen and gcollapse weights are meant to work. I've changed the title to one I think is more precise but LMK if you disagree. Cheers.

from stata-gtools.

@erciomunoz This is not a gtools bug per se, so to the extent it's a bug at all it would be on Stata. As documented in the help files, weights in gegen and gcollapse are meant to mimic collapse (see weights section of the collapse help pdf). You can see

clear
set obs 5
g units=1
g weight=5
collapse (sum) units [w = weight]
disp units

Gives 5 as the answer. The reason is that aweights are normalized to sum to the number of observations. Try fweight, pweight, or iweight to see the difference. Also

clear
set obs 5
g units=_n
g weight=_n
gegen total =total(units) [w=weight]
gegen totalu=total(units)
sum total*

Gives a different answer, showing weights are not being ignored (they just work different than how you might intuit). Cheers.

from stata-gtools.

Thanks for the quick reply.

This makes total sense, now I regret a little bit that I did not spend more time thinking about it before posting.

from stata-gtools.