treatment effect IPW, RA, AIPW, ...

[josef...@gmail.com]

I picked up another one of my 2015 PRs

Treatment effects under conditional independence, selection based on observables

uses GMM under the hood to compute inference (following Stata's approach)

https://github.com/statsmodels/statsmodels/pull/8034

should be ready to merge in a few days.

for docs, see stata manual for `teffects`

Josef

example based on unit test case

https://gist.github.com/josef-pkt/8a8755d2930cb23f7282cf683232bf17

teff = TreatmentEffect(mod, tind, results_select=res_probit)

for m, _ in methods:
    res = getattr(teff, m)()
    print("\n", m)
    print(res.summary_frame())

 ra
             coef    std err           z         P>|z|  Conf. Int. Low  Conf. Int. Upp.
ATE   -239.639211  23.824021  -10.058722  8.408247e-24     -286.333435      -192.944988
POM0  3403.242272   9.525207  357.288006  0.000000e+00     3384.573209      3421.911335
POM1  3163.603060  21.863509  144.697867  0.000000e+00     3120.751371      3206.454750

 ipw
             coef    std err           z         P>|z|  Conf. Int. Low  Conf. Int. Upp.
ATE   -230.688638  25.816745   -8.935621  4.048989e-19     -281.288528      -180.088748
POM0  3403.462709   9.571412  355.586256  0.000000e+00     3384.703085      3422.222332
POM1  3172.774071  24.001048  132.193145  0.000000e+00     3125.732880      3219.815261

 aipw
             coef    std err           z         P>|z|  Conf. Int. Low  Conf. Int. Upp.
ATE   -230.989648  26.214445   -8.811541  1.234375e-18     -282.369017      -179.610280
POM0  3403.355674   9.568514  355.682783  0.000000e+00     3384.601731      3422.109616
POM1  3172.366025  24.427402  129.869153  0.000000e+00     3124.489197      3220.242854

 aipw_wls
             coef    std err           z         P>|z|  Conf. Int. Low  Conf. Int. Upp.
ATE   -227.195618  27.372036   -8.300282  1.038645e-16     -280.843822      -173.547414
POM0  3403.250651   9.596571  354.631943  0.000000e+00     3384.441717      3422.059585
POM1  3176.055033  25.654642  123.800406  0.000000e+00     3125.772859      3226.337206

 ipw_ra
             coef    std err           z         P>|z|  Conf. Int. Low  Conf. Int. Upp.
ATE   -229.967078  26.629411   -8.635830  5.830196e-18     -282.159765      -177.774391
POM0  3403.335639   9.571288  355.577620  0.000000e+00     3384.576260      3422.095018
POM1  3173.368561  24.871955  127.588224  0.000000e+00     3124.620425      3222.116697