答：在跑过movestay之后，用mspredict计算：tt=参与者处理效应=工人参与private项目的事实-工人不参与private项目的反事实； tu=非参与者处理效应=非工人不参与private项目的事实-非工人参与private项目的反事实；

summarize tt tu //输出att,atu

ate为att和atu的加权平均

g te2=tt

replace te2=tu if te2==.

summarize tt te2 tu //输出att,ate,atu及其标准误

tw (kdensity 工人参与private项目的事实)(kdensity 工人不参与private项目的反事实)

.....

movestay之后计算att, ate, atu还是比较少的，毕竟它的原理与PSM不同。有人指出：

The subtle error comes from omitting the negative sign for the definition of the Mills ratio for the non‐participants group. Thus, even if the results of movestayStata command are accurate, some of those of the mspredict are wrong. At this stage, I have corrected temporarily this post command (mspredict_ar.ado) and I will contact the movestayauthors. In addition, I have programmed for the PEP teams the post command msat to estimate the treatment effects.

显然movestay authors没有搭理他！

As you didn't give an accessible reference for the paper I imagine that people didn't bother to search for it. Did the author state that it was installed on SSC? Because, as you say, no such program is there.

此人再也没有公布他的msat.ado,似乎蒸发了。