Looping over gender and total sample

[Christa Hansen]

Hi,

I have a typical sex variable where 1 is male and 2 is female. I am trying to build a value 3 which would be everyone, so that I might have a loop that goes from male to female, to everyone.

I am really struggling for some reason to build number 3. I thought this would be simple but it is not proving to be. Does anyone have any suggestions?

This is what I have now but its not working.

gen allsex = .
egen allsex1 = max(sex == 1 | sex == 2)
replace allsex = 3 if allsex1 == 1
replace allsex = 1 if sex == 1
replace allsex = 2 if sex == 2

[Mike Lacy]

You cannot have what you want, as it would require a person to have more than one value on the sex variable. (I think you misunderstand how variables work in a program like Stata.) Fortunately, what you want is not something necessary or useful. I'd guess that you want to run some command (say -summarize-) separately for each gender and for the whole sample. There are many ways to do this. Here are several, to demonstrate techniques:

// Preferred approach.
summarize YourVariable   // regardless of the value of sex
// Each gender separately.  See -help by-, a powerful feature of Stata.
bysort sex: summarize YourVariable
//
// Another way
summarize YourVariable
summarize YourVariable if sex == 1
summarize YourVariable if sex == 2

A loop could be devised to do what you want, and might in some circumstance be useful if your variable, unlike sex, had many different values rather than just a few. Here's that technique, which is not desirable here:

// Another way
summarize YourVariable
levelsof sex, local(sexvalues)
foreach v of local sexvalues {
    summarize YourVaraible if sex == `v'
}

[Christa Hansen]

Hi, the request was specific to me wanting to run the loop for male, female, total sample, which is specific to the needs of my model, and trying to save a bunch of copying and pasting. Not so much the need for it to be all in one variable (though that would have been handy for a loop). I do realise in what I suggested, value 3 would have been a function of 1 and 2. I just wasn't sure if there was a way around this.

Thanks.

[Sonnen Blume]

Hi, the request was specific to me wanting to run the loop for male, female, total sample, which is specific to the needs of my model, and trying to save a bunch of copying and pasting. Not so much the need for it to be all in one variable (though that would have been handy for a loop). I do realise in what I suggested, value 3 would have been a function of 1 and 2. I just wasn't sure if there was a way around this.

Thanks.

A binary variable cannnot have three categories. You could just run three seperate sets of analyses for male, female and pooled sample.