Announcement

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

  • Looking to plot a quaternary dependent variable with ordered logistic regression

    Hello,
    I'm looking make some kind of useful plot based on ordered logistic regression results. Specifically, I have:
    • Students' ratings of a midterm's fairness (Likert scale, completely unfair - completely fair, 4 total options)
    • Students' grades (percentage grades)
    I have regression parameters from:
    > ologit fair grade

    Is there a way to meaningfully plot these results? I know how to plot this kind of thing with a binary outcome, but not with 4 possible outcomes.

    Any advice is greatly appreciated! I've googled my heart out, but haven't found a solution yet.
    Thanks,
    Mike
    Using Stata v15.1 on Win7

  • #2
    Sure. After the -ologit- run
    Code:
    margins, at(grade = (0(10)100))
    marginsplot
    You will get a graph with four curves on it, each representing the probability of one of the four Likert-like responses as a function of grade.

    Note: Since you said that the grades are percentage grades, I chose to run the plot from a grade of 0 to 100. But perhaps the lowest achieved score was bigger than 0 or the highest achieved score less than 100. In that case you might replace the (0(10)100) by some more appropriate numbers that span, and fill with reasonable density, the range of achieved scores.

    Also, -marginsplot- accepts most -graph twoway- options, so you can customize the appearance of your graph.

    Comment


    • #3
      That worked beautifully! Thank you for your answer and explanation.

      Comment


      • #4
        Although it's inverting response and predictor, a dotplot of grade by rating would show all the information in the data. See dotplot or stripplot (SSC).

        Comment

        Working...
        X