Compute a mean index only for cases having at least 3 valid values over 4 variables

[Amelie Nickel]

Hey everyone, I am using Stata Version 14.2. I want to compute a mean index over 4 variables but only count cases having at least 3 non missing values on those 4 variables. In SPSS you can specifiy this by setting the valid case number(3)

compute Marktf = mean.3 (emp05, emp06, ok02, ok01).

Is there a way to transfer this SPSS code to Stata?

[Emil Alnor]

use the min option in the alpha command. See

help alpha

[Rich Goldstein]

maybe I don't understand what a "mean index" is, but I don't see that -alpha- does what I think it means; try this (assuming that all variables are present within an observation):

egen numthere=rownonmiss(emp05 emp06 ok02 ok01)
egen meanindex=rowmean(emp05 emp06 ok02 ok01) if numthere>=3

[Emil Alnor]

maybe I don't understand what a "mean index" is, but I don't see that -alpha- does what I think it means; try this (assuming that all variables are present within an observation):

egen numthere=rownonmiss(emp05 emp06 ok02 ok01)
egen meanindex=rowmean(emp05 emp06 ok02 ok01) if numthere>=3

Alpha produces the same variable as your egen commands, but also has the possibility of producing some usable index statistics like Cronbach's Alpha or item-rest-correlation:

sysuse auto, clear
alpha length turn displacement weight, min(3) gen(index1) item

egen numthere=rownonmiss(length turn displacement weight)
egen index2=rowmean(length turn displacement weight) if numthere>=3

pwcorr index*
browse index*

[Amelie Nickel]

Thank you so much Emil Alnor and Rich Goldstein!

Both codes you shared actually produced the same accurate new variable.

I didn't know that the 'alpha command' could be used to generate new variables. Thanks for enlightening me.

Regarding the 'mean index', what I meant was that I intended to create a new variable by computing the mean. I apologize for any confusion I may have caused.