Announcement

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

  • How to interpret the marginal effects for dummy variables in logit regression

    I ran a logit model, lhs is risk tolerance (=1 if respondent willing to take high risk; =0 otherwise), rhs are demographic factors such as age, generation, year, ... then I used mfx command to show marginal effect. For example, the marginal effect for generation Y was -0.0166503* (5% significant), then my interpretation was "Each respondents belonging to the Y generation decreased the likelihood of reporting risk tolerance by 1.66 per cent." I was told that my interpretation was wrong but I couldn't manage to find where.

  • #2
    Almost:
    "...decreased the likelihood of reporting risk tolerance by 1.66 percentage points"

    I am not sure about the command you use, the standard is nowadays
    Code:
    logit...
    margins, dydx(*)
    Best wishes

    (Stata 16.1 MP)

    Comment


    • #3
      It is also worth pointing out that whether or not you include all dummy categories changes your interpretation. If one generation is omitted, then you interpret your effect relative to the omitted baseline. Otherwise, it is interpreted as Felix #2 says.

      Comment


      • #4
        Felix Bittmann Thank you for your help, I will consider changing the margin command
        Chris Boudreaux Yes yes, that's also my struggle. I collected data of respondents from 5 generations (Y, X, Silent, Baby Boomer and GI) and included them all to Stata by generating 1 dummy variable for each generation. Then I used "logit ...." command and Stata automatically omitted 1 generation (GI). Could you be more specific on the "reference category", like how should I interpret the effect relative to the omitted baseline?

        Comment


        • #5
          Sure. The interpretation is "Each respondent belonging to the Y generation decreased the likelihood of reporting risk tolerance by 1.66 percentage points compared to respondents belonging to GI."

          Comment


          • #6
            Chris Boudreaux Hi Chris, it's me again. I just updated my data and it showed insignificant results on every generation (except GI as reference category) so is it correct if I interpret it like "generations do not affect risk tolerance" or should I relate it to the ommitted baseline?

            Comment


            • #7
              The latter interpretation--it should be interpreted relative to the omitted category.

              Comment

              Working...
              X