Displaying rho after biprobit, margins in the output table

[Olga Goldfine]

Dear Statalist members, I would be very thankful for an advice. I am running a biprobit model, and use eststo/esttab commands to output marginal effects. I need to add the rho coefficient (ideally, with significe level) to the output, but so far I was not able to figure out how to do it. When I use the estadd scalar corr = e(rho), the resulting scalar is empty after I run the eststo: margins
Here is the code I have worked so far with:

biprobit y1 y2 female $controls_fainc i.date , vce(robust)
estadd scalar corr = e(rho)
eststo: margins, predict (pmarg1) dydx(female $controls_fainc) atmeans noatlegend post
sum y1 if e(sample)==1
estadd scalar mean= r(mean)
eststo: margins, predict (pmarg2) dydx(female $controls_fainc) atmeans noatlegend post
sum y2 if e(sample)==1
estadd scalar mean= r(mean)


esttab using "${results}\table1.tex", replace style(tex) keep(female $controls_fainc) order(female $controls_fainc) ///
nonumber star(* 0.10 ** 0.05 *** 0.01) varlabels(_cons Constant) ///
cells(b(star fmt(3)) se(par fmt(3))) label booktabs alignment(D{.}{.}{-1}) collabels(none) nodepvar ///
stats(mean N corr , layout("@" "@" "\multicolumn{2}{c}{@}") labels( "\text{Mean dependent variable}" "\text{N observations}" "\text{Rho}") fmt(3 0 3))


Would be very obliged for any tip! Thank you!
Olga

[Andrew Musau]

estout is from the Stata Journal (FAQ Advice #12). Notice that your first estimation results pertain to the biprobit, and the -post- option of margins overwrites these estimates. Therefore, you need to store these estimates in a local and add them after the margins command.

biprobit y1 y2 female $controls_fainc i.date , vce(robust)
local corr = e(rho)
eststo: margins, predict (pmarg1) dydx(female $controls_fainc) atmeans noatlegend post
estadd scalar corr = `corr'
sum y1 if e(sample)==1

I need to add the rho coefficient (ideally, with significe level)

You can either add the p-value as a separate scalar below the coefficient or you can add the coefficient with significance starts to indicate the significance level. esttab does not make adding stars to scalars (or stats) easy (unless it is 3 stars). However, I provide code to do this in #2 of the following link:

https://www.statalist.org/forums/for...gression-table

In your case, you want first to save the P-value in a local and specify the conditions, e.g., as

estadd local corr "`= cond(`pval'<0.01,"`:di %5.3f `=`corr'''***", ...