Issue with Missing Value Check for Different Variable Types

[KADER-YETTEFTI Sami]

Hello,

I'm facing an issue with tagging observations with missing values in a dataset containing both numerical and string variables. Variables:

• _304753 (double)
• _304754 (double)
• _304755 (long)
• _306872 (str14)
• _309592 (double)
• _900420 (long)
• _900565 (long)

I need to tag observations as "void" if any of these variables have missing values. Here is my code: Code Used for Transformation

gen void = 0
local bob _304753 _304754 _304755 _309592 _900420 _900565
foreach var in `bob' {
    replace void = 1 if missing(`var')
}
replace void = 1 if _306872 == ""

Diagnostics Attempted

• Using missing() function:

replace void = 1 if missing(_304753) & missing(_304754) & missing(_304755) & _306872 == "" & missing(_309592) & missing(_900420) & missing(_900565)

Result: No missing values considered missing.

• Using == . for numerical variables:

replace void = 1 if _304753 == . & _304754 == . & _304755 == . & _306872 == "" & _309592 == . & _900420 == . & _900565 == .

• Result: Syntax error.

I need to correctly tag observations where all specified variables are missing to remove duplicates and keep valid information for analysis. Any suggestions?

Thank you!

Best regards,
Sami

[Clyde Schechter]

I can't tell what you are trying to do here. Your first attempted code, using a loop, will mark void = 1 if any of those variables as a missing value. The other two will mark void = 1 if all of those variables have missing values. And on my setup, using a made up data set that meets your description, all three of them run without any error messages.

I will also note that thee is no need to exclude _306872 from the loop in the first one: it will work just fine even though _306872 is string and the others are numeric. Stata has no problem with that.

Also, for the second and third attempts, it is not necessary to start with -gen void = 0- and then -replace void = 1...- You can just do

gen void = (_304753 == . & _304754 == . & _304755 == . & _306872 == "" ///
//     & _309592 == . & _900420 == . & _900565 == .)

(and analogously for the version using the -missing()- function.)

So I think you just need to be clear on whether you want void = 1 when any of the variables is missing or when all of them are missing and then just go with one of these existing attempts.

[Nick Cox]

See also

help egen

for its rowmiss() and rownonmiss() functions.