楼主看看工具变量法的原理，直接用笨办法做，估计可以的。看看伍德里奇的书。

没有做过

你可以下载cmp命令
看到其中的例子有，这需要你自己看命令和帮助了

These examples go beyond standard commands in Stata 12:

    . webuse laborsup

    * Regress an unbounded, continuous variable on an instrumented, binary one. 2SLS is consistent but less efficient.
    . cmp (other_inc = fem_work) (fem_work = kids), ind($cmp_cont $cmp_probit) qui robust
    . ivreg other_inc (fem_work = kids), robust

    * Now regress it on a left-censored one, female income, which is only modeled for observations in which the woman works.
    . gen byte ind2 = cond(fem_work, cond(fem_inc, $cmp_cont, $cmp_left), $cmp_out)
    . cmp (other_inc=fem_inc kids) (fem_inc=fem_edu), ind($cmp_cont ind2)

    * "IV-oprobit"
    . cmp (kids = fem_educ) (fem_educ = fem_work), ind($cmp_oprobit $cmp_cont) tech(dfp) nolr
    . margins, dydx(*) predict(eq(#1) pr outcome(#2)) force

    * Ordered probit with Heckman selection modeling
    . preserve
    . webuse womenwk, clear
    . gen selectvar = wage<.
    . gen wage3 = (wage > 10)+(wage > 30) if wage < .
    . cmp (wage3 = education age) (selectvar = married children education age), ind(selectvar*$cmp_oprobit $cmp_probit) qui
    . restore

有序probit的工具变量回归也适用这条命令吗？

也同样适用！

请问楼主运用成功了吗，最近也在找这个命令