Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Create an index

    I need to group 3 variables to build an index in Stata. Each variable refers to a question in a questionnaire. The point is that they all have different scales: one variable has a scale from 1 to 7 (1 = never, 7 = always, another variable from 1=low to 3=high, and the third 1 to 2 (it's a yes or no question). How do I build an index with these three variables?

  • #2
    Constanca:
    welcome to this forum.
    I would take a look at -pca- entry in Stata. pdf manual.
    Kind regards,
    Carlo
    (Stata 19.0)

    Comment


    • #3
      The same question was posted on Stack Overflow, answered and then the thread was deleted, which -- while legal behaviour there -- was disappointing. This is more or less what I said there in addition to some puzzlement expressed by myself and one other person.

      There are many indexes possible here but no guarantee that any is much use.

      You could go

      Code:
      gen wanted = (question1 - 1) / 6 + (question2 - 1) / 2 + (question3 - 1)
      which yields an index between 0 and 3 with the provisos that

      1. If you want a mean between 0 and 1 then divide by 3

      2. If any meaning is reversed compared with the others, then reverse its scale.

      3. While equal weighting is the simplest procedure, it is arbitrary otherwise.

      4. If these are predictors in a regression, then I would not seek a composite. I would use all of these as they come, with some thought given to whether they are crude measures or to be treated as factor variables.

      I am less hopeful than perhaps is Carlo Lazzaro of the utility of PCA here. While unfazed by different ranges, PCA treats all these variables as if counted or measured, i.e. as on interval scale.
      Last edited by Nick Cox; 29 Nov 2021, 06:46.

      Comment


      • #4
        Nick:
        I suggested -pca- on the brink of desperation, remembering myself dealing with a similar issue before my Stata conversion.
        Your reply, strange as it may seem and luckily for the original poster, ouperforms my previous one.
        Kind regards,
        Carlo
        (Stata 19.0)

        Comment

        Working...
        X