求大神赐教：Heckman two stage 第一步selection model 有3组以上response， for example, 0, 1, 2，如果用multinomial logit, 如何计算多个inverse mills ratio？求code，求推荐相关文献或者教材，感激不尽！


原始两步法code，第一步是probit, dummy有0,1两个response，求大神在此基础上修改或者指导，如果此code有问题也请指出：


first stage:

proc logistic data=Dataset1;
class dummy;
model dummy = X1 X2 X3
/link=glogit;
output out=Dataset2 xbeta=xbeta stdxbeta=stdxbeta;
run;
quit;

data dataset2; set dataset2;
IMR=pdf('normal',xbeta)/cdf('normal',xbeta);
run;

second stage:

proc reg data=dataset2;
model Y= dummy X4 X5 X6 IMR
;ods graphics off;
run;
quit;