looping by a specific variable

[Mo Hos]

Hi everyone.

I generated the following

egen firmlevel=group(CompanyID AnnualReportYear)

And then I tried to use the loop command as follows:

foreach v of varlist Main* {
gen sum`v'=sum(`v'), by (firmlevel)

However, I received the following message:

option by() not allowed

Does anyone has an idea how I can do it?

[Carlo Lazzaro]

Mo:
try, instead:

foreach v of varlist Main* {
bysort firmlevel: gen sum`v'=sum(`v')
 }

[Nick Cox]

I have to wonder whether you really want egen, total()

A cumulative or running sum doesn't make obvious sense here.

[Mo Hos]

Mo:
try, instead:

foreach v of varlist Main* {
bysort firmlevel: gen sum`v'=sum(`v')
}

Thanks. It worked.

[Mo Hos]

I have to wonder whether you really want egen, total()

A cumulative or running sum doesn't make obvious sense here.

I have a dummy variable for each variable of the Main* (e.g. Main_gender=1 if female, 0 otherwise) and I would like to generate the sum of each variable for each firm-year observation (the sum of 1s in my case). Therefore, I will have the number of, for example, females for each fimr-year observation.

In short: I would do the following code for each variable:

egen totalvar=sum(var), by (firmlevel)

[Mo Hos]

egen firmlevel=group(CompanyID AnnualReportYear)

I realised that there are different results generated from the following two methods:

For the first method, I would do this for each variable:

egen totalvar=sum(var), by (firmlevel)

For the second one, I would use the following command:

foreach v of varlist Main* {
bysort firmlevel: gen Sum`v'=sum(`v')
}

The results of the first method are the one I want. However, I would like to do it like the second method.

[Nick Cox]

 
 egen totalvar = sum(var), by (firmlevel)

is perfectly legal, but as from Stata 9 the sum() function of egen is undocumented. The name total() for sums, meaning totals, is meant to insert a slim difference given that the Stata function sum() returns a cumulative or running sum. Indeed Mata disambiguates thoroughly by using the name runningsum().

A bridge between the two is

 
 bysort firmlevel: gen Sum`v' = sum(`v')  by firmlevel: replace Sum`v' = Sum`v'[_N]

as the last cumulative sum is the total concerned. Example, the cumulative sum of 1, 2, 3 runs 1, 3, 6 and 6 is also the total.

[Mo Hos]

@Nick Cox Worked perfectly. Thanks.

[Nick Cox]

For anyone puzzled by #7 the second code block contains two statements

bysort firmlevel: gen Sum`v' = sum(`v')

by firmlevel: replace Sum`v' = Sum`v'[_N]