1

Im looking for marginal effects in my bivariate probit model in R. For the univariate case via

model <- glm(formula, family=binomal(link="probit"))

its clear how to get them: simply take

margins(model)

But now I have a model estimated with the gjrm package. My example looks like:

library(Zelig)
data("sanction")
liste_biprob <- list(import ~ coop + cost, export ~ cost + target)
mr <- c("probit", "probit")
biprobit_1 <- gjrm(liste_biprob, data=sanction, Model="B", margins=mr)

From this model I want to extract the marginal effects now. margins() doesnt work anymore. Also I tried to split my output in a way that margins could work, but I failed. Im would be very grateful if anybody has a way to calculate the marginal effects for my case!

r_10_b
  • 11
  • 1
  • What does it mean if you say it "doesn't work anymore"? Do you have any error messages? – 0x1C1B Aug 03 '21 at 17:55
  • margins(data=sanction, biprobit_1, variables = "cost") produces: Fehler in terms.default(model) : no terms component nor attribute – r_10_b Aug 03 '21 at 18:42
  • I also figured out that this works, but its not satisfying: dydx(data=sanction, biprobit_1, variable = "cost", eq=1) It gives the marginal effects and with mean() its possible to calculate the AME, but there is no information on the standard error (SE) or the z-value or the confidence intervals like you get usually when using margins(). – r_10_b Aug 03 '21 at 18:44

0 Answers0