egen by(id) returning zero for missing values

[juliana pinto]

Dear Statalists, please, could anyone help me?
I have the data set example below and am trying to use egen to calculate by group, but the results for obs with missing values are returning 0. How can I fix this?
Many thanks.

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(id x1 x2 num den totalnum totalden total)
1  .  3         .  .3333333         0  .4242424         0
2 17 10       1.7        .1 2.7555556 .15555556 17.714285
3  .  .         .         .       1.2       .05        24
1  . 11         .  .0909091         0  .4242424         0
2 19 18 1.0555556 .05555556 2.7555556 .15555556 17.714285
3 24 20       1.2       .05       1.2       .05        24
end

gen num = x1/x2

gen den = 1/x2

egen totalnum = total(num), by(id)

egen totalden = total(den), by(id)
gen total = totalnum/totalden

[juliana pinto]

.

[Andrew Musau]

egen has a missing option:

bys id: egen totalnum = total(num), missing

[Nick Cox]

Stata's logic was that if you want to regard (say) the sum of 1, 2, 3 as 6 and the sum of 1, 2, 3, and missing as 6, and the sum of 1, 2, 3, missing and missing as 6, and so on, then the rule you are using is that the sum of missings is (is regarded as) zero. Note that Stata uses this rule in summarize in returning r(sum) and in the function sum() which also ignores missings.

As said, if you don't want to do that, egen has a documented option to insist that the sum of one or more missings is itself reported as missing.

[juliana pinto]

Thank you both very much for the solution and explanations!