I recently received help from the forum that helped me obtain the marginal effects I was after. Now I am a bit stumped on interpretation. I have the following AMEs after an interaction, where reduccat is respondent's education and keducat is the highest level of education of their children. I see that I have on significant dy/dx. How can this be substantively interpreted? My main IV is child's education, and I want to know if this association is moderated by the respondent's (their parents's) education. Since 2.k keducat, where reduccat is "HS," does this mean that the effect of having a child with a college degree is stronger for respondent's with a high school degree? Or is it "The effect of having a child with a college degree, compared to a child with HS, is amplified when the respondent has only a HS degree?" I am not sure how to compare to the omitted keduc category.
I took this approach based on the categorical by categorical approach presented here:
https://stats.oarc.ucla.edu/stata/se...ions-stata/#s5
I took this approach based on the categorical by categorical approach presented here:
https://stats.oarc.ucla.edu/stata/se...ions-stata/#s5
Code:
. mimrgns reduccat, dydx(keducat) note: option predict() not specified; predict(xb) assumed Multiple-imputation estimates Imputations = 20 Average marginal effects Number of obs = 5,046 Average RVI = 0.0062 Largest FMI = 0.0111 DF adjustment: Large sample DF: min = 155,667.77 avg = 3980951.47 Within VCE type: Delta-method max = 1.18e+07 Expression : Linear prediction (log odds), predict(xb) dy/dx w.r.t. : 1.keducat 2.keducat ------------------------------------------------------------------------------ | dy/dx Std. err. t P>|t| [95% conf. interval] -------------+---------------------------------------------------------------- 0.keducat | (base outcome) -------------+---------------------------------------------------------------- 1.keducat | reduccat | HS | .0907721 .1525444 0.60 0.552 -.2082106 .3897548 Some Col | -.1920499 .2541508 -0.76 0.450 -.6901776 .3060779 Col+ | -.4071653 .4889296 -0.83 0.405 -1.36545 .5511192 -------------+---------------------------------------------------------------- 2.keducat | reduccat | HS | .5117982 .1500929 3.41 0.001 .2176191 .8059772 Some Col | .3979326 .2460594 1.62 0.106 -.0843371 .8802024 Col+ | .0360967 .4631199 0.08 0.938 -.8716017 .9437951 ------------------------------------------------------------------------------ Note: dy/dx for factor levels is the discrete change from the base level.