library(car); library(lmtest) #para desvios-padrao robustos
#estimar MPL
linprob <- lm(inlf~ nwifeinc+educ+exper+I(exper^2)+age+kidslt6+kidsge6,data=MROZ)
#estimar modelo logit
logitres<-glm(inlf~ nwifeinc+educ+exper+I(exper^2)+age+kidslt6+kidsge6,family=binomial(link=logit),data=MROZ)
summary(logitres)
#estimar modelo probit
probitres<-glm(inlf~ nwifeinc+educ+exper+I(exper^2)+age+kidslt6+kidsge6,family=binomial(link=probit),data=MROZ)
summary(probitres)
#calcular indices lineares nos valores individuais
xb.log<-predict(logitres)
xb.prob<-predict(probitres)
#fatores APE=media(g(xb))
factor.log<-mean(dlogis(xb.log))
factor.prob<-mean(dnorm(xb.prob))
cbind(factor.log, factor.prob)
#efeitos parciais médios=beta*factor
APE.lin<-coef(linprob)*1
APE.log<-coef(logitres)*factor.log
APE.prob<-coef(probitres)*factor.prob
#tabela APEs
cbind(APE.lin, APE.log,APE.prob)