Sum scores to create a scale

[Magda Karyda]

Hi all,
I need some help in the right choice of commands in the following . I have 15 binary variables and 14,000 observations. A score of 1 is assigned to each item answered “no”, 0 to items answered “yes" and . for missing items. I am trying to sum the scores to create a scale.
Any suggestions you might have are more than welcome,
Thank you in advance
Magda

[Joseph Coveney]

Look into the egen command. You'll probably want to include an if condition for excluding sum-scores for respondents who don't have your threshold number of nonmissing responses.

[Magda Karyda]

Thank you, I will look into that.

[Magda Karyda]

Thank you again, I think it works

[wbuchanan]

Depending on what the items are and the purpose for creating the scale you may want to consider using IRT/CTT to generate a scale that will be adjusted for measurement error. The sum score itself could be very noisy, in which case subsequent analyses would be biased by the error to some degree.