Marginal effects of a multivariate probit model

[Lily Ksh]

Dear Statalisters,

I am struggling to estimate the marginal effects for my MVP model. I have gone through the previous posts (eg: https://www.statalist.org/forums/for...variate-probit). However, I am unable to run these codes effectively.

My model has 5 dependent variables within the multivariate model. Currently my codes look smth like:

mvprobit (y1 x1 x2....x15) (y2 x1 x2....x15) (y3 x1 x2...x15) (y4 x1 x2....x15) (y5 x1 x2....x15)

Can someone please help me with coding and estimating the marginal effects of the explanatory variables in my MVP model?

Thank you for your help.

[George Ford]

Took this from:

https://www.researchgate.net/post/How_to_estimate_Marginal_effects_of_Multivariate_Probit_model_using_STATA

I'm not sure this is legit, but it works. Used biprobit to check.

clear

set obs 1000

g x1 = rnormal()
g x2 = rnormal()

g y1 = 2*x1 + 1*x2 + rnormal() > 0.5
g y2 = -2*x1 + 2*x2 + rnormal() > 0.5

mvprobit (y1 x1 x2) (y2 x1 x2)
mvppred pred_xb, xb
gen coef_x1_1 = _b[y1:x1]
gen coef_x2_1 = _b[y1:x2]
gen coef_x1_2 = _b[y2:x1]
gen coef_x2_2 = _b[y2:x2]

gen ME_x1_1 = normalden(pred_xb1)*coef_x1_1
gen ME_x2_1 = normalden(pred_xb1)*coef_x2_1
gen ME_x1_2 = normalden(pred_xb2)*coef_x1_2
gen ME_x2_2 = normalden(pred_xb2)*coef_x2_2

summ ME_*

biprobit (y1 x1 x2) (y2 x1 x2)
margins, dydx(*) predict(pmarg1)
margins, dydx(*) predict(pmarg2)

Running them individually will give you the same margins, I think.

[Lily Ksh]

Hi George, thank you so much!