Announcement

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

  • create new variable from likert scale

    Hi
    I have many likert scale responses from my survey in 3-4 different categories of subjects (e.g. environmental attitudes, general attitudes, certain behaviours, etc). I want to sort of reduce the information into variables, so that I can include them in a regression model.
    Any suggestion will help.
    Thanks

  • #2
    Well, if multiple items measure the same constructs, you can use alpha
    Code:
    alpha var1 var2 var3, gen(construct)
    help alpha
    If you are not sure whether they load on the same constructs, you can run a factor analysis or PCA
    Code:
    help pca
    Or you can include them unchanged easily using factor variable notation
    Code:
    regress y i.var1 i.var2
    Best wishes

    Stata 18.0 MP | ORCID | Google Scholar

    Comment


    • #3
      Thanks a lot !

      Comment


      • #4
        Originally posted by Felix Bittmann View Post
        Well, if multiple items measure the same constructs, you can use alpha
        Code:
        alpha var1 var2 var3, gen(construct)
        help alpha
        Hi sorry for asking seeking out help again.

        i got a scale reliability coeff of 0.1870 but unsure how to interpret it? it seems I have a very low reliability coefficient...

        alpha ipol pol pets inher tech, gen(construct)

        Test scale = mean(unstandardized items)
        Reversed item: pets

        Average interitem covariance: .0934579
        Number of items in the scale: 5
        Scale reliability coefficient: 0.1870


        Comment


        • #5
          Yes this is very low, results below 0.6 are critical. This means that either the variables do not measure the same construct. But without knowing the data in depth I cannot give more detailed advice. I suggest proceeding with a PCA to check whether there are multiple dimensions in the items.
          Best wishes

          Stata 18.0 MP | ORCID | Google Scholar

          Comment


          • #6
            Thanks

            the variables represent interest (1=None at all, 2=A little, 3=A moderate amount, 4=A lot, 5=A great deal, 6=Don't Know):
            ipol:international politics
            pol:international politics
            pets:pets
            inher:leaving inheritance
            tech:technology to improve lives

            Comment


            • #7
              First, make sure to set value 6 to a missing, otherwise the results will be quite useless (replace VAR = . if VAR == 6)
              Second, I do not understand what construct you attempt to create. What is the common theme of politics and pets?
              Best wishes

              Stata 18.0 MP | ORCID | Google Scholar

              Comment

              Working...
              X