use ma_car,replace /*Heckman Correction, Y=car7, X=dualrole, InstrumentVariable=toplender */
probit dualrole toplender relsize hostile competing allcash crossind toehold tenderoffer pubdeal acq_size acq_q acq_leverage acq_roa runup volatility toptierib
estimates store selectionbias_1
predict Z,xb
gen invmills=normalden(Z)/normal(Z)
ivreg2 car7 dualrole $dealx $acqx $industry $yr invmills,cluster(gvkey)
estimates store selectionbias_2
跑出来的结果如下 invmill不显著 但X依然显著 说明不存在selection bias是吗?