Replace missing data with mean or average for summative scale purposes

[Lis Flores]

Hi, I fully understand the issues with mean imputation, but I still need to use this as a quick solution to build proxy scales (not for publishable work). I have 65 variables and ~10k observations. Observations are named rf1, rf2, rf3...rf65. I've tried this:

foreach var of varlist rf* {
2. quietly sum 'var'
3. replace 'var' = r(mean) if missing('var')
4. }

but I get the "invalid name r(198);" error message. How else could I approach this?

[Daniel Schaefer]

Hi Lis, I just have a quick sanity-check question before we really dig in - you use the back tick character below the tilde symbol (`) at the start of a local, right? From the code you provide here it looks like you surround your locals in single quotes - it should however be `var' not 'var', correct? Or does the full error look like this:

' invalid name
r(198);

[Daniel Schaefer]

Otherwise, your solution seems to work with a very basic test case:

clear
input float(rf1 rf2 rf3 rf4 rf5)
1 1 5 1 3
2 2 . . .
3 3 3 . .
4 . 2 . .
5 5 1 1 .
end

foreach var of varlist rf* {
    quietly sum `var'
    replace `var' = r(mean) if missing(`var')
}

list, clean noob

. list, clean noob

    rf1    rf2    rf3   rf4   rf5  
      1      1      5     1     3  
      2      2   2.75     1     3  
      3      3      3     1     3  
      4   2.75      2     1     3  
      5      5      1     1     3

[Lis Flores]

You were right Daniel, I was the black tick mistake that kept my code from running correctly. Thank you so much, problem solved!