Announcement

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

  • likert scale analysis

    I have collected data from 300 respondents on their perception about water stress in their area. The responses have been captured using 5 point likert scale. How should I proceed further for analysis on stata??

  • #2
    Pragya, the question is too vague to answer easily. You said a 5-point Likert scale - do you mean you have one question that's a 5-point Likert item, or do you mean you have several 5-point Likert items that form a scale?

    If you are using a validated scale, you might consider just proceeding with a regression analysis. If you have one 5-point Likert item, you would be best advised to use ordinal logistic or probit regression (little to no substantive difference between the two). If you have a scale, most people would sum the items up and use a linear regression. That's the simplest approach. There is valid criticism of this approach, however.

    If you are not using a validated scale, then you may wish to think about doing some basic checks. You could run Cronbach's Alpha (help alpha), but this statistic is also not without flaws. You could run an exploratory factor analysis (help factor) to check if your scale is consistent with a one-factor solution. There are more complex ways to go about showing construct validity.
    Please use the code delimiters to show code and results - use the # button on the formatting toolbar, between the " (double quote) and <> buttons.

    Please use the command -dataex- to show a representative sample of data; it is installed already if you have Stata 14.2 or 15.1, else you can install it by typing

    Code:
    ssc install dataex

    Comment


    • #3
      I have almost 10 questions on 5 pt.likert scale. I summed them up using the syntax gen index = (var1 + var2.....varn). Then when I use the tab command i get range of 9 to 38. Now how to proceed further?? I read that you need to deduct 5 or 10 to get better scores..what is that?

      Comment


      • #4
        Subtracting a constant is a linear transformation, so that does not change anything substantive. So that is not a priority.

        It is still unclear what you want to do with the index you created. You are asking us what your next step should be, but that obviously depends on the direction you want to go. without that there is nothing we can do.
        ---------------------------------
        Maarten L. Buis
        University of Konstanz
        Department of history and sociology
        box 40
        78457 Konstanz
        Germany
        http://www.maartenbuis.nl
        ---------------------------------

        Comment


        • #5
          Sir, Im unable to understand as to how to use these scores obtained after summation for analysis

          waterstress Freq. Percent Cum.

          19 1 0.33 0.33
          20 3 1.00 1.33
          21 3 1.00 2.33
          22 12 4.00 6.33
          23 17 5.67 12.00
          24 37 12.33 24.33
          25 57 19.00 43.33
          26 44 14.67 58.00
          27 32 10.67 68.67
          28 31 10.33 79.00
          29 24 8.00 87.00
          30 19 6.33 93.33
          31 9 3.00 96.33
          32 7 2.33 98.67
          33 2 0.67 99.33
          39 2 0.67 100.00

          Total 300 100.00

          This is the output. Now please tell how to interpret it and which statistical tools can be used further

          Comment


          • #6
            Originally posted by Pragya Sharma View Post
            I have almost 10 questions on 5 pt.likert scale. I summed them up using the syntax gen index = (var1 + var2.....varn). Then when I use the tab command i get range of 9 to 38. Now how to proceed further?? I read that you need to deduct 5 or 10 to get better scores..what is that?
            As Maarten said, if you don't clearly describe what you want to do, it's hard for us to guess.

            I would urge you to be a bit more precise. "Almost 10 questions" is a bit of an odd way to say 9 questions, if you meant that. Also, if each question were scored 1, 2, 3, 4, or 5, and you had 9 questions, your total possible score range should be 9 to 45. Is this what you have, except that the actual scores range from 9 to 38?

            You can deduct 1 point from each question, e.g.

            Code:
            foreach v of varlist var? {
            replace `v' = `v' - 1
            }
            egen index = rowtotal(var?)
            It's a personal preference of mine, but it doesn't change anything substantive as Maarten said. You do not need to deduct 5 or 10 points.

            As I said earlier, you may be justified in conducting exploratory or confirmatory factor analysis to verify that your instrument has the factor structure it's supposed to. Otherwise, you could just use the total score in a regression.

            If you need any more detailed answers, please ask a specific question. Otherwise, it is impossible to help you.
            Please use the code delimiters to show code and results - use the # button on the formatting toolbar, between the " (double quote) and <> buttons.

            Please use the command -dataex- to show a representative sample of data; it is installed already if you have Stata 14.2 or 15.1, else you can install it by typing

            Code:
            ssc install dataex

            Comment


            • #7
              Thankyou for your help. I think I can solve the problem now

              Comment


              • #8
                Sincerely, and please don't take it amiss, it always strikes me as odd when one starts the collection of data involving human beings, but the way to proceed with "further analysis" is yet to be unveiled.

                This process should be done the other way, and the study design is previously writtten for due purpose.

                Additionally, just summing up scores may not be the best approach. Among several reasons, doing something "new" in terms of fiddling with the scores will prompt to a necessary validation, hence hard work ahead.

                Therefore, a great starting point is to check the instrument and how the authors recommend to use it.

                That said, item response theory might well be the analysis you're looking for.

                Hopefully that helps.
                Best regards,

                Marcos

                Comment

                Working...
                X