Announcement

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

  • The item sign in Cronbach's Alpha result

    Good morning, I met a problem when doing the Cronbach's Alpha analysis, so I really need help.
    In my original data set, items like FA1 and FA2 were reverse-coded Likert-type questions. I have manually reverse code these items using excel, labeled as FA1r and FA2r. The problem is, after performing multiple imputations on both data sets and running alpha FA1 FA2 FA3 FA4 FA5, it was expected that the “sign” of FA3 will be “-“ since all the other 4 items were reversely coded and vice versa, running alpha FA1r FA2r FA3 FA4r FA5r all the “sign” will be “+” since they have been corrected by me. But this is what actually happened. In spite of being reversed, the "sign" FA3 is still negative.

  • #2
    This implies that the FA3 item's responses are negatively correlated with the responses to the other reverse-coded variables.

    You can force Stata to do the calculation preserving the sign for FA3 by adding the -asis- option to your command. The resulting alpha will be different, and may well be low because you are, in effect, forcing Stata to include a variable whose actual responses (not the response set offered) go in the opposite direction of the other items. A better idea is to to check that the data are actually correct recordings of the responses, and, if they are, to try to understand why the way people responded to item FA3 is opposite to the way they responded to the other items.

    Comment


    • #3
      I'd also recode in Stata rather than excel. That way you have documentation and may detect an error.

      Comment

      Working...
      X