Summary statistics for variables with multiple responses

[Akanimo Enang]

Hello,

I am currently having some challenges computing some variables which allow the respondent to input multiple responses. I am working with data obtained from a questionnaire filled by nearly 4,000 respondents.

I think this my challenge is best explained using an example. Say for instance, a question asked "What social media networks are you registered with?". In theory, a respondent could select any combination of Twitter, Facebook, Snapchat etc. Right now, I am only able to sum up the number of respondents who use any given network with the egen command on STATA (e.g 1000 use twitter, 2000 use facebook etc). For the purpose of my analysis, it would be far more effective and insightful to see the actual overlaps between usage of different networks. In essence, I'm trying to distill to a level where I can make informed deductions like, 500 respondents use both Facebook and Snapchat while 300 use Facebook, Twitter & Pinterest etc.

I would greatly appreciate any assistance/advice anyone is able to provide.

[Richard Williams]

I think you could use the group function of egen. Then each combination would have a unique value, e.g. facebook only would have one value, facebook + twitter would have another value, etc,

[Nick Cox]

There is a lot of advice about this e.g.

SJ-5-1  st0082  . . . . . . . . . . . . . . . Tabulation of multiple responses
        (help _mrsvmat, mrgraph, mrtab if installed)  . . . . . . . .  B. Jann
        Q1/05   SJ 5(1):92--122
        introduces new commands for the computation of one- and
        two-way tables of multiple responses


SJ-3-1  pr0008   Speaking Stata: On structure & shape: the case of mult. resp.
        . . . . . . . . . . . . . . . . . . . . . . . .  N. J. Cox & U. Kohler
        Q1/03   SJ 3(1):81--99                                   (no commands)
        discussion of data manipulations for multiple response data