比如，通过计算P-score进行匹配，可以为每个观测值生成PS_weight.

在估算Treatment效应时，该权重如何嵌入到分析中去？即其加权在被解释变量上，还是加权在每个解释变量上？还是同时加权到解释变量与被解释变量中？

即采用  Y*weight=a1*x1+a2*x2.....?

还是采用  Y=a1*x1*weight+a2*x2*weight?

还是采用  Y*weight=a1*x1*weight+a2*x2*weight?