麻烦问一下各位老师同学，做了似不相关双边probit模型后，怎样预测对应的反米尔斯比率，目的是把两个方程的反米尔斯放到第二阶段的回归中，希望各位老师同学解惑，谢谢大家！

//help biprobit

//Examples

//webuse school ,giving up for firewall

cd D:\\biprobit_heckprobit_mill

use school,clear

//Bivariate probit regression

biprobit private vote logptax loginc years

//Seemingly unrelated bivariate probit regression

biprobit (private = logptax loginc years) (vote = logptax years)

//Seemingly unrelated bivariate probit regression with robust standard errors

biprobit (private = logptax loginc years) (vote = logptax years), vce(robust)

****************************help heckprobit******************************************

//[R] heckprobit -- Probit model with sample selection

//(View complete PDF manual entry)

//Fit a probit model with sample selection

heckprobit private years logptax, sel(vote=years loginc logptax)

//Replay results, but display legend of coefficients rather than the statistics for the coefficients

heckprobit, coeflegend

//stata中heckprobit命令的查询可知rho即为实际的逆米尔兹比

//https://bbs.pinggu.org/thread-2660588-1-1.html

****probit 提取inverse mills ratio

probit private vote logptax loginc years,nolog // possibility of being open regionpredict gw, xbg lambda=normalden(gw)/normal(gw) if private==1

replace lambda=-normalden(gw)/normal(-gw) if private==0 // generate inverse mills ratio (selection model)

//怎么在stata中计算mills ratio

probit private vote logptax loginc years,nolog

predict gw, xb

replace lambda=normalden(gw,0,1)/normal(gw) if private==1

replace lambda=-normalden(gw,0,1)/normal(-gw) if private==0

*y是0-1变量，x1-xn是解释变量，lambda即IMR

//http://bbs.pinggu.org/thread-1087688-1-1.html?winzoom=1