Announcement

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

  • Advice on how to interpret margins and marginsplot (interaction in linear model)

    Dear Statslist

    I am unsure how to interpret the margins and marginsplot with regards to my interaction in a linear model. More specifically I would like to know, how it shows in marginsplot when my interaction is not significant - is it because the confidence intervals overlap?

    According to my regressionoutput the interaction between direktionsmodel and sanktionsregime is not significant but how can is this showed in the margins and marginsplot?













    Thanks
    Ann





    Attached Files

  • #2
    More specifically I would like to know, how it shows in marginsplot when my interaction is not significant - is it because the confidence intervals overlap?
    No, and there is no way to extract this information from -marginsplot-. The test of significance of your interaction is in the regression output. Marginsplot shows you the effect each level of direktionsmodel and their confidence intervals, but the overlapping (or not) of the confidence intervals does not tell you whether the interaction is statistically significant. If the confidence intervals do not overlap, then the difference is statistically non-significant. But it does not work the other way around: you can have a statistically significant interaction even though the confidence intervals overlap. You might want to take a look at https://www.cscu.cornell.edu/news/statnews/stnews73.pdf, which illustrates the situation in a different context, but the principle is no different.

    Comment


    • #3
      Thank you very much for the response it was very helpful.

      I have a follow up question. If the confidence interval include 0 in marginsplo,t does this mean that the difference is statistically insignifcant because of the margins null hypothesis?

      Attached Files

      Comment


      • #4



        I forgot to upload my regressions output for reference for my question about confidence intervals including 0 in margins

        Thank you in advance

        Click image for larger version

Name:	output1nr.2.png
Views:	1
Size:	302.3 KB
ID:	1457843

        Click image for larger version

Name:	output2nr.2.png
Views:	1
Size:	172.4 KB
ID:	1457844
        Click image for larger version

Name:	output3.pnnr.2g.png
Views:	1
Size:	56.0 KB
ID:	1457845


        Comment


        • #5
          If the confidence interval include 0 in marginsplo,t does this mean that the difference is statistically insignifcant because of the margins null hypothesis?
          So, in the example you show, the confidence interval for the marginal effect of direktionsmodel when santkionsregime takes on its maximum value includes 0, and, yes, you can infer from that that that marginal effect is not statistically significant. (Which, however, should not be interpreted as meaning there is no effect. Statistically not significant means only that the data identify the effect in an imprecise way so that we cannot be certain whether the effect is positive or negative or zero.)

          Comment


          • #6
            Dear Clyde

            Thank you so much for all your help and your comments about the meaning of significance.

            Best regards
            Ann

            Comment

            Working...
            X