------------------------------------------------------------------------------- log: c:\bill\spring2011\jtoa_training.log log type: text opened on: 11 Apr 2011, 14:43:41 . * open log file; . * get some measure of partial compliance; . tab treatment assignment, row column; +-------------------+ | Key | |-------------------| | frequency | | row percentage | | column percentage | +-------------------+ =1 if | repondend | went into | job | =1 if assigned to job training, | training, =0 =0 | otherwise otherwise | 0 1 | Total -----------+----------------------+---------- 0 | 1,684 1,282 | 2,966 | 56.78 43.22 | 100.00 | 98.88 37.72 | 58.13 -----------+----------------------+---------- 1 | 19 2,117 | 2,136 | 0.89 99.11 | 100.00 | 1.12 62.28 | 41.87 -----------+----------------------+---------- Total | 1,703 3,399 | 5,102 | 33.38 66.62 | 100.00 | 100.00 100.00 | 100.00 . * define the exogenous covariates in the regressions; . local xlist hsorged black hispanic age* > wkless13 married; . * get OLS estimates of the impact of job training; . reg earnings `xlist' treatment; Source | SS df MS Number of obs = 5102 -------------+------------------------------ F( 11, 5090) = 44.07 Model | 1.6937e+11 11 1.5398e+10 Prob > F = 0.0000 Residual | 1.7783e+12 5090 349363638 R-squared = 0.0870 -------------+------------------------------ Adj R-squared = 0.0850 Total | 1.9476e+12 5101 381814524 Root MSE = 18691 ------------------------------------------------------------------------------ earnings | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- hsorged | 3827.904 592.5323 6.46 0.000 2666.286 4989.523 black | -2609.299 629.53 -4.14 0.000 -3843.449 -1375.15 hispanic | 145.7383 907.6409 0.16 0.872 -1633.628 1925.105 age2225 | 5834.095 1546.975 3.77 0.000 2801.359 8866.83 age2629 | 7052.115 1565.75 4.50 0.000 3982.572 10121.66 age3035 | 5608.143 1542.558 3.64 0.000 2584.066 8632.22 age3664 | 4541.626 1555.574 2.92 0.004 1492.031 7591.221 age4554 | 2143.938 1705.458 1.26 0.209 -1199.494 5487.37 wkless13 | -6571.3 573.6429 -11.46 0.000 -7695.886 -5446.713 married | 6539.927 583.9488 11.20 0.000 5395.136 7684.718 treatment | 3574.821 531.9044 6.72 0.000 2532.059 4617.582 _cons | 10702.56 1579.97 6.77 0.000 7605.138 13799.98 ------------------------------------------------------------------------------ . * use assignment as instrument for treatment; . * estimates 2sls model -- ignores DGP for treatment; . reg earnings `xlist' treatment (`xlist' assignment); Instrumental variables (2SLS) regression Source | SS df MS Number of obs = 5102 -------------+------------------------------ F( 11, 5090) = 40.14 Model | 1.6458e+11 11 1.4962e+10 Prob > F = 0.0000 Residual | 1.7831e+12 5090 350305477 R-squared = 0.0845 -------------+------------------------------ Adj R-squared = 0.0825 Total | 1.9476e+12 5101 381814524 Root MSE = 18716 ------------------------------------------------------------------------------ earnings | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- treatment | 1604.478 910.2379 1.76 0.078 -179.9802 3388.935 black | -2567.936 630.5684 -4.07 0.000 -3804.121 -1331.751 hispanic | 237.2353 909.5097 0.26 0.794 -1545.795 2020.266 age2225 | 5813.21 1549.078 3.75 0.000 2776.35 8850.069 age2629 | 6954.926 1568.282 4.43 0.000 3880.419 10029.43 age3035 | 5543.021 1544.829 3.59 0.000 2514.493 8571.55 age3664 | 4429.381 1558.237 2.84 0.004 1374.566 7484.196 age4554 | 1997.164 1708.641 1.17 0.243 -1352.506 5346.835 wkless13 | -6566.331 574.4186 -11.43 0.000 -7692.438 -5440.223 married | 6636.138 585.8452 11.33 0.000 5487.629 7784.647 hsorged | 3909.796 594.1231 6.58 0.000 2745.059 5074.533 _cons | 11489.81 1609.352 7.14 0.000 8334.79 14644.83 ------------------------------------------------------------------------------ . * show first stage probit for treatment effect; . probit treatment `xlist' assignment; Iteration 0: log likelihood = -3468.6232 Iteration 1: log likelihood = -2451.7414 Iteration 2: log likelihood = -2354.1117 Iteration 3: log likelihood = -2341.312 Iteration 4: log likelihood = -2340.618 Iteration 5: log likelihood = -2340.6141 Probit regression Number of obs = 5102 LR chi2(11) = 2256.02 Prob > chi2 = 0.0000 Log likelihood = -2340.6141 Pseudo R2 = 0.3252 ------------------------------------------------------------------------------ treatment | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hsorged | .155783 .0480717 3.24 0.001 .0615642 .2500018 black | .0727836 .0512158 1.42 0.155 -.0275974 .1731647 hispanic | .1388195 .0739482 1.88 0.060 -.0061163 .2837553 age2225 | -.0453221 .1249173 -0.36 0.717 -.2901555 .1995114 age2629 | -.2219305 .1260843 -1.76 0.078 -.4690513 .0251903 age3035 | -.1158453 .1243525 -0.93 0.352 -.3595718 .1278811 age3664 | -.216024 .1257496 -1.72 0.086 -.4624888 .0304407 age4554 | -.2236721 .1384802 -1.62 0.106 -.4950884 .0477442 wkless13 | -.0189245 .0466568 -0.41 0.685 -.1103702 .0725211 married | .1196308 .0474253 2.52 0.012 .0266789 .2125828 assignment | 2.611109 .0899701 29.02 0.000 2.434771 2.787447 _cons | -2.327078 .1509532 -15.42 0.000 -2.622941 -2.031215 ------------------------------------------------------------------------------ . * the mismeasured treatment effect model; . * the syntax is treatreg y x, treat(t=x z) where y is the outcome; . * x is the list of exogenous factrors, t is the treatment variable; . * and z are the instruments; . treatreg earnings `xlist', treat(treatment=`xlist' assignment); Iteration 0: log likelihood = -59753.761 Iteration 1: log likelihood = -59753.747 Iteration 2: log likelihood = -59753.747 Treatment-effects model -- MLE Number of obs = 5102 Wald chi2(11) = 444.86 Log likelihood = -59753.747 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- earnings | hsorged | 3894.686 593.0899 6.57 0.000 2732.251 5057.121 black | -2575.568 629.5268 -4.09 0.000 -3809.418 -1341.718 hispanic | 220.3526 907.9759 0.24 0.808 -1559.248 1999.953 age2225 | 5817.063 1546.557 3.76 0.000 2785.867 8848.26 age2629 | 6972.859 1565.696 4.45 0.000 3904.152 10041.57 age3035 | 5555.038 1542.3 3.60 0.000 2532.185 8577.89 age3664 | 4450.092 1555.655 2.86 0.004 1401.064 7499.12 age4554 | 2024.247 1705.787 1.19 0.235 -1319.035 5367.528 wkless13 | -6567.248 573.4842 -11.45 0.000 -7691.256 -5443.239 married | 6618.385 584.7987 11.32 0.000 5472.201 7764.57 treatment | 1968.04 883.0844 2.23 0.026 237.2265 3698.854 _cons | 11344.55 1604.448 7.07 0.000 8199.89 14489.21 -------------+---------------------------------------------------------------- treatment | hsorged | .1574098 .0480538 3.28 0.001 .063226 .2515936 black | .0718658 .0511572 1.40 0.160 -.0284005 .1721321 hispanic | .1380297 .0739335 1.87 0.062 -.0068773 .2829367 age2225 | -.0500384 .1247828 -0.40 0.688 -.2946083 .1945315 age2629 | -.2251408 .1259115 -1.79 0.074 -.4719227 .0216411 age3035 | -.1189649 .1242019 -0.96 0.338 -.3623962 .1244664 age3664 | -.2201547 .1255955 -1.75 0.080 -.4663174 .026008 age4554 | -.2216769 .1383267 -1.60 0.109 -.4927922 .0494384 wkless13 | -.0183814 .0466375 -0.39 0.693 -.1097892 .0730264 married | .1175322 .0474124 2.48 0.013 .0246056 .2104587 assignment | 2.607888 .0896832 29.08 0.000 2.432112 2.783664 _cons | -2.320411 .1506683 -15.40 0.000 -2.615716 -2.025107 -------------+---------------------------------------------------------------- /athrho | .0799418 .0351193 2.28 0.023 .0111093 .1487743 /lnsigma | 9.83553 .0099483 988.67 0.000 9.816032 9.855029 -------------+---------------------------------------------------------------- rho | .079772 .0348958 .0111089 .1476863 sigma | 18686.01 185.8936 18325.2 19053.93 lambda | 1490.62 653.7324 209.3276 2771.912 ------------------------------------------------------------------------------ LR test of indep. eqns. (rho = 0): chi2(1) = 5.19 Prob > chi2 = 0.0227 ------------------------------------------------------------------------------ . * now estimate a model that is identified solely based on the; . * non-linearities in the model; . treatreg earnings `xlist', treat(treatment=`xlist'); Iteration 0: log likelihood = -63324.767 (not concave) Iteration 1: log likelihood = -61583.156 Iteration 2: log likelihood = -61318.037 Iteration 3: log likelihood = -61126.462 Iteration 4: log likelihood = -60871.661 (not concave) Iteration 5: log likelihood = -60868.491 (not concave) Iteration 6: log likelihood = -60868.297 (not concave) Iteration 7: log likelihood = -60867.95 (not concave) Iteration 8: log likelihood = -60867.665 (not concave) Iteration 9: log likelihood = -60867.518 (not concave) Iteration 10: log likelihood = -60867.379 (not concave) Iteration 11: log likelihood = -60867.24 (not concave) Iteration 12: log likelihood = -60867.085 (not concave) Iteration 13: log likelihood = -60866.825 (not concave) Iteration 14: log likelihood = -60866.448 (not concave) Iteration 15: log likelihood = -60865.974 (not concave) Iteration 16: log likelihood = -60865.763 (not concave) Iteration 17: log likelihood = -60865.541 (not concave) Iteration 18: log likelihood = -60865.304 (not concave) Iteration 19: log likelihood = -60865.055 (not concave) Iteration 20: log likelihood = -60864.796 (not concave) Iteration 21: log likelihood = -60864.525 (not concave) Iteration 22: log likelihood = -60864.242 (not concave) Iteration 23: log likelihood = -60863.946 (not concave) Iteration 24: log likelihood = -60863.639 (not concave) Iteration 25: log likelihood = -60863.319 (not concave) Iteration 26: log likelihood = -60862.988 (not concave) Iteration 27: log likelihood = -60862.645 (not concave) Iteration 28: log likelihood = -60862.291 (not concave) Iteration 29: log likelihood = -60861.927 (not concave) Iteration 30: log likelihood = -60861.553 (not concave) Iteration 31: log likelihood = -60861.17 (not concave) Iteration 32: log likelihood = -60860.779 (not concave) Iteration 33: log likelihood = -60860.381 (not concave) Iteration 34: log likelihood = -60859.978 (not concave) Iteration 35: log likelihood = -60859.571 (not concave) Iteration 36: log likelihood = -60859.161 (not concave) Iteration 37: log likelihood = -60858.749 (not concave) Iteration 38: log likelihood = -60858.338 Iteration 39: log likelihood = -60850.815 Iteration 40: log likelihood = -60849.761 Iteration 41: log likelihood = -60849.761 Treatment-effects model -- MLE Number of obs = 5102 Wald chi2(11) = 657.74 Log likelihood = -60849.761 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- earnings | hsorged | 2784.881 711.8934 3.91 0.000 1389.595 4180.166 black | -3136.126 754.4108 -4.16 0.000 -4614.744 -1657.508 hispanic | -1019.619 1088.794 -0.94 0.349 -3153.615 1114.378 age2225 | 6100.07 1852.519 3.29 0.001 2469.199 9730.941 age2629 | 8289.941 1876.199 4.42 0.000 4612.658 11967.22 age3035 | 6437.54 1847.744 3.48 0.000 2816.028 10059.05 age3664 | 5971.208 1864.444 3.20 0.001 2316.966 9625.451 age4554 | 4013.294 2044.868 1.96 0.050 5.425945 8021.162 wkless13 | -6634.589 686.931 -9.66 0.000 -7980.949 -5288.229 married | 5314.524 702.5547 7.56 0.000 3937.542 6691.506 treatment | 28670.2 1529.987 18.74 0.000 25671.48 31668.92 _cons | 675.6925 1971.925 0.34 0.732 -3189.209 4540.594 -------------+---------------------------------------------------------------- treatment | hsorged | .0951809 .0395507 2.41 0.016 .017663 .1726988 black | .0480052 .0421552 1.14 0.255 -.0346175 .1306279 hispanic | .1133994 .0601488 1.89 0.059 -.0044901 .2312889 age2225 | -.0307723 .1035774 -0.30 0.766 -.2337802 .1722356 age2629 | -.146704 .1050382 -1.40 0.163 -.352575 .0591671 age3035 | -.0851166 .1033332 -0.82 0.410 -.2876459 .1174127 age3664 | -.1520699 .1042509 -1.46 0.145 -.3563979 .052258 age4554 | -.1850599 .1143236 -1.62 0.106 -.4091301 .0390103 wkless13 | .0201202 .0382599 0.53 0.599 -.0548679 .0951082 married | .1168591 .0389197 3.00 0.003 .0405778 .1931404 _cons | -.2365043 .1048147 -2.26 0.024 -.4419373 -.0310713 -------------+---------------------------------------------------------------- /athrho | -.870412 .054429 -15.99 0.000 -.9770909 -.7637331 /lnsigma | 10.01603 .0210262 476.36 0.000 9.974815 10.05724 -------------+---------------------------------------------------------------- rho | -.7015834 .027638 -.7518038 -.6432706 sigma | 22382.3 470.6146 21478.66 23323.96 lambda | -15703.05 922.6357 -17511.38 -13894.72 ------------------------------------------------------------------------------ LR test of indep. eqns. (rho = 0): chi2(1) = 40.73 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ . log close; log: c:\bill\spring2011\jtoa_training.log log type: text closed on: 11 Apr 2011, 14:43:58 -------------------------------------------------------------------------------