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

  • Using Cronbach's alpha vs Krippendorf's alpha

    Hi stata users,

    I have a question regarding of which type of reliability test (Cronbach's or Krippendorf's) use for my data. I'm working in a manuscript in which we use the value orientantions approach for environmental concern and behaviour (egoism, altruism, biocentrism). Each value orientation are composed of 4 or 5 items (Psycometric variables, de Groot and Steg's value orientations ) and the respondents have to rate each item according to his/her grade of agreement (1 to 9 scale).

    In all the literature we've found in this topic, the cronbach's alpha has being used to test the internal consistency of value orientations. But someone suggested us the krippendorf alpha to test inter-rater reliability. I was wondering if you know about Krippendorf's alpha test and if it is pertinent for value orientantions measurement.

    With the Cronbach's alpha we find acceptable consistency but low alpha with Krippendorf's test.

    Thanks in advance for your comments,


  • #2
    Cronbach's alpha assesses the internal consistency of a unidimensional scale while Krippendorff's alpha assesses inter-rater agreement corrected for chance agreement. From what you describe, I would guess you are interested in the former but only you can ultimately know that. You should probably check the dimensionality of the scales that you are using since Cronbach's alpha only applies to unidimensional scales. It also assumes tau-equivalence but that is another story.

    Concerning the reported differences, you do not show any data, you do not say which commands you have used, you do not show syntax, you do not show any output and you do not give precise numbers; consequentially, I cannot really comment on this. However, since both measures are mathematically differently defined and assess conceptually different things, we would not necessarily expect that the results agree.