Survey Subscale descriptive stats

[Chris Laugen]

Hello, I am trying to create basic descriptive stats for a survey that I administered. The survey uses a likert scale with five response options (1,2,3,4,5) for strongly disagree (1) - strongly agree (5). The items are organized into subscales so that each item has three to four sub-items (i.e. 6a, 6b, 6c). I am interested in obtaining some basic descriptive stats based on these sub-scales in addition to per individual item. So for example, I would like the mean, sd, number of missing, etc. for 6a, 6b, and 6c combined. If I am interested in the mean across a sub-scale (i.e. rq6a rq6b rq6c) instead of just one item (i.e. rq6a), is the code below the most efficient or is there a better way to get this result? I am using Stata 11 and the sample below (first 10 observations) is how my data are organized.

input byte(rq6a rq6b rq6c)
2 4 4
5 4 5
2 2 2
4 4 4
5 4 4
3 4 4
3 3 3
5 5 5
4 3 4
3 3 3
end

egen rq6mean = rowmean(rq6a rq6b rq6c)
sum rq6mean

Thanks for your help.

[Clyde Schechter]

That's close to the way I would do it. There is one caveat here. How will you handle missing responses? In most settings, we would not score a 3-item subscale if any of the items has a missing response, and there are very few situations in which one would score a 3-item subscale when 2 of the items have missing responses. So, assuming you want to exclude observations with any missing responses on these three items from scoring it would be:

egen rq6mean = rowmean(rq6a rq6b rq6c) if !missing(rq6a, rq6b, rq6c)

[Chris Laugen]

Thank you for your response and the additional information on missing responses. I noticed that these commands will still calculate a row mean even if there is one missing response. I think it would be more prudent to remove the missing cases with the suggestion you made.