I have a proportion DV. I can't use the
Code:
fracreg
Code:
xtgee y i.x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12, family(binomial 1) link(probit) corr(exchangeable) vce(robust)
I found the link on fractional models (https://www.stata.com/stata14/fracti...utcome-models/) and dyex is used for margins. I tried to use dyex, but I couldn't, because it says factor variables not allowed with dyex. So, I decided to use
Code:
margins, dydx(*) margins x1
Code:
margins x1 Predictive margins Number of obs = 13,405 Model VCE : Semirobust Expression : Pr(y != 0), predict() ------------------------------------------------------------------------------ | Delta-method | Margin Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- x1 | 0 | .2681136 .0049269 54.42 0.000 .2584571 .2777702 1 | .3810374 .0503208 7.57 0.000 .2824105 .4796644 ------------------------------------------------------------------------------
Does this read that when x goes from 0 to 1 the proportion (y) increases by 11.3%? I found Richard Williams's Slides https://www3.nd.edu/~rwilliam/stats/Margins01.pdf. Are average marginal effects the same thing as average partial effects?
Lastly, I'm obviously not well-versed in econometrics, but is there another way to interpret the results of x1? With the margin command (EDIT: Or is it because of the regression?), if I interpreted correctly, it reads from 0 to 1 (for lack of better words). Is there a way to read the results as when x=1 (yes) or when x=0 (no) not as "when no goes to yes." (Again, sorry for the layman's terms. I gladly welcome better terminology).
Thank you.
0 Response to Understanding Margins with a Fractional Proportion DV
Post a Comment