------------------------------------------------------------------------------- name: log: /Users/carlos/GitHub/starter-academic-v501/content/post/stata_matc > hing/analysis.log log type: text opened on: 29 Apr 2026, 12:22:33 . . di _newline(2) . di "================================================================" ================================================================ . di " Treatment Effects in Stata: Six Estimators" Treatment Effects in Stata: Six Estimators . di " Maternal Smoking and Birth Weight Case Study" Maternal Smoking and Birth Weight Case Study . di "================================================================" ================================================================ . . *--------------------------------------------------- . * 1. Data loading and exploration . *--------------------------------------------------- . . di _newline(2) "=== 1. DATA LOADING AND EXPLORATION ===" === 1. DATA LOADING AND EXPLORATION === . . use "https://github.com/quarcs-lab/data-open/raw/master/ametrics/cattaneo2.dt > a", clear (Excerpt from Cattaneo (2010) Journal of Econometrics 155: 138–154) . . * Describe the variables we will use . describe bweight mbsmoke mage mmarried fage medu prenatal1 fbaby Variable Storage Display Value name type format label Variable label ------------------------------------------------------------------------------- bweight int %9.0g Infant birthweight (grams) mbsmoke byte %9.0g mbsmoke 1 if mother smoked mage byte %9.0g Mother's age mmarried byte %11.0g mmarried 1 if mother married fage byte %9.0g Father's age medu byte %9.0g Mother's education attainment prenatal1 byte %9.0g YesNo 1 if first prenatal visit in 1 trimester fbaby byte %9.0g YesNo 1 if first baby . . * A peek at the data . list bweight mbsmoke mmarried mage fage medu prenatal1 fbaby in 40/49, noobs +---------------------------------------------------------------------------+ | bweight mbsmoke mmarried mage fage medu prenat~1 fbaby | |---------------------------------------------------------------------------| | 2790 Nonsmoker Married 29 30 17 Yes Yes | | 4082 Nonsmoker Married 30 31 12 Yes Yes | | 3089 Nonsmoker Married 27 37 17 Yes No | | 2580 Smoker Not married 18 0 10 No Yes | | 3714 Nonsmoker Married 35 33 14 Yes No | |---------------------------------------------------------------------------| | 4196 Nonsmoker Married 35 35 16 Yes No | | 2523 Nonsmoker Not married 16 19 9 Yes Yes | | 2693 Smoker Not married 25 28 12 Yes No | | 3204 Nonsmoker Married 27 33 16 Yes Yes | | 3572 Smoker Not married 19 20 12 Yes Yes | +---------------------------------------------------------------------------+ . . * Summary statistics . summarize bweight mbsmoke mage mmarried fage medu prenatal1 fbaby Variable | Obs Mean Std. dev. Min Max -------------+--------------------------------------------------------- bweight | 4,642 3361.68 578.8196 340 5500 mbsmoke | 4,642 .1861267 .3892508 0 1 mage | 4,642 26.50452 5.619026 13 45 mmarried | 4,642 .6996984 .4584385 0 1 fage | 4,642 27.26713 9.354411 0 60 -------------+--------------------------------------------------------- medu | 4,642 12.68957 2.520661 0 17 prenatal1 | 4,642 .8013787 .3990052 0 1 fbaby | 4,642 .4379578 .4961893 0 1 . . * Treatment prevalence . tab mbsmoke 1 if mother | smoked | Freq. Percent Cum. ------------+----------------------------------- Nonsmoker | 3,778 81.39 81.39 Smoker | 864 18.61 100.00 ------------+----------------------------------- Total | 4,642 100.00 . . * Sample size by treatment . tab mbsmoke, summarize(bweight) | Summary of Infant birthweight 1 if mother | (grams) smoked | Mean Std. dev. Freq. ------------+------------------------------------ Nonsmoker | 3412.9116 570.68711 3,778 Smoker | 3137.6597 560.89305 864 ------------+------------------------------------ Total | 3361.6799 578.81962 4,642 . . *--------------------------------------------------- . * 2. Naive comparison and descriptive plot . *--------------------------------------------------- . . di _newline(2) "=== 2. NAIVE COMPARISON (UNADJUSTED) ===" === 2. NAIVE COMPARISON (UNADJUSTED) === . . * Naive difference in means: this is the BIASED comparison . regress bweight mbsmoke, vce(robust) Linear regression Number of obs = 4,642 F(1, 4640) = 168.33 Prob > F = 0.0000 R-squared = 0.0343 Root MSE = 568.88 ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. t P>|t| [95% conf. interval] -------------+---------------------------------------------------------------- mbsmoke | -275.2519 21.21501 -12.97 0.000 -316.8434 -233.6604 _cons | 3412.912 9.285455 367.55 0.000 3394.708 3431.115 ------------------------------------------------------------------------------ . estimates store te_naive . . * Save the naive estimate so we can compare later . scalar naive_b = _b[mbsmoke] . scalar naive_se = _se[mbsmoke] . di "Naive (unadjusted) gap: " %7.2f naive_b " grams (SE = " %5.2f naive_se ") > " Naive (unadjusted) gap: -275.25 grams (SE = 21.22) . . * Figure 1: kernel density of bweight by smoking status . twoway /// > (kdensity bweight if mbsmoke==0, lcolor("106 155 204") lwidth(medthick)) > /// > (kdensity bweight if mbsmoke==1, lcolor("217 119 87") lwidth(medthick)) > /// > , /// > title("Birth Weight by Maternal Smoking Status", size(medium)) /// > subtitle("Raw, unadjusted distributions", size(small)) /// > xtitle("Infant birth weight (grams)") ytitle("Density") /// > legend(order(1 "Non-smokers" 2 "Smokers") position(6) rows(1)) /// > graphregion(color(white)) plotregion(color(white)) /// > note("Smokers' distribution is shifted left of non-smokers' --- but is th > is causal?", size(vsmall)) . graph export "stata_matching_density_bweight.png", replace width(2400) file stata_matching_density_bweight.png written in PNG format . . *--------------------------------------------------- . * 3. Method 1: Regression Adjustment (RA) . *--------------------------------------------------- . . di _newline(2) "=== 3. METHOD 1: REGRESSION ADJUSTMENT (RA) ===" === 3. METHOD 1: REGRESSION ADJUSTMENT (RA) === . . * teffects ra: potential-outcome means, ATE, ATT . teffects ra (bweight mmarried mage prenatal1 fbaby) (mbsmoke), pomeans nolog Treatment-effects estimation Number of obs = 4,642 Estimator : regression adjustment Outcome model : linear Treatment model: none ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- POmeans | mbsmoke | Nonsmoker | 3403.242 9.525207 357.29 0.000 3384.573 3421.911 Smoker | 3163.603 21.86351 144.70 0.000 3120.751 3206.455 ------------------------------------------------------------------------------ . teffects ra (bweight mmarried mage prenatal1 fbaby) (mbsmoke), ate nolog Treatment-effects estimation Number of obs = 4,642 Estimator : regression adjustment Outcome model : linear Treatment model: none ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATE | mbsmoke | (Smoker | vs | Nonsmoker) | -239.6392 23.82402 -10.06 0.000 -286.3334 -192.945 -------------+---------------------------------------------------------------- POmean | mbsmoke | Nonsmoker | 3403.242 9.525207 357.29 0.000 3384.573 3421.911 ------------------------------------------------------------------------------ . estimates store te_ra . teffects ra (bweight mmarried mage prenatal1 fbaby) (mbsmoke), atet nolog Treatment-effects estimation Number of obs = 4,642 Estimator : regression adjustment Outcome model : linear Treatment model: none ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATET | mbsmoke | (Smoker | vs | Nonsmoker) | -223.3017 22.7422 -9.82 0.000 -267.8755 -178.7278 -------------+---------------------------------------------------------------- POmean | mbsmoke | Nonsmoker | 3360.961 12.75749 263.45 0.000 3335.957 3385.966 ------------------------------------------------------------------------------ . estimates store te_ra_att . . * Manual recreation: regress separately for D=0 and D=1, then average . preserve . di _newline "--- Manual recreation of RA ---" --- Manual recreation of RA --- . regress bweight mmarried mage prenatal1 fbaby if mbsmoke==0 Source | SS df MS Number of obs = 3,778 -------------+---------------------------------- F(4, 3773) = 30.00 Model | 37916015 4 9479003.74 Prob > F = 0.0000 Residual | 1.1922e+09 3,773 315979.752 R-squared = 0.0308 -------------+---------------------------------- Adj R-squared = 0.0298 Total | 1.2301e+09 3,777 325683.776 Root MSE = 562.12 ------------------------------------------------------------------------------ bweight | Coefficient Std. err. t P>|t| [95% conf. interval] -------------+---------------------------------------------------------------- mmarried | 160.9513 24.55483 6.55 0.000 112.8092 209.0933 mage | 2.546828 1.935504 1.32 0.188 -1.247908 6.341564 prenatal1 | 64.40859 25.83668 2.49 0.013 13.75338 115.0638 fbaby | -71.3286 19.4895 -3.66 0.000 -109.5396 -33.11763 _cons | 3202.746 51.20076 62.55 0.000 3102.362 3303.13 ------------------------------------------------------------------------------ . predict y0_hat, xb . regress bweight mmarried mage prenatal1 fbaby if mbsmoke==1 Source | SS df MS Number of obs = 864 -------------+---------------------------------- F(4, 859) = 3.69 Model | 4586519.21 4 1146629.8 Prob > F = 0.0055 Residual | 266914157 859 310726.609 R-squared = 0.0169 -------------+---------------------------------- Adj R-squared = 0.0123 Total | 271500676 863 314601.015 Root MSE = 557.43 ------------------------------------------------------------------------------ bweight | Coefficient Std. err. t P>|t| [95% conf. interval] -------------+---------------------------------------------------------------- mmarried | 133.6617 41.33023 3.23 0.001 52.54164 214.7818 mage | -7.370881 3.983409 -1.85 0.065 -15.18924 .4474735 prenatal1 | 25.11133 43.64639 0.58 0.565 -60.55473 110.7774 fbaby | 41.43991 41.82146 0.99 0.322 -40.6443 123.5241 _cons | 3227.169 102.3392 31.53 0.000 3026.305 3428.033 ------------------------------------------------------------------------------ . predict y1_hat, xb . generate te_i = y1_hat - y0_hat . summarize te_i Variable | Obs Mean Std. dev. Min Max -------------+--------------------------------------------------------- te_i | 4,642 -239.6392 99.008 -488.4602 8.261719 . di "Manual RA estimate of ATE: " %7.2f r(mean) " grams" Manual RA estimate of ATE: -239.64 grams . restore . . *--------------------------------------------------- . * 4. Method 2: Inverse-Probability Weighting (IPW) . *--------------------------------------------------- . . di _newline(2) "=== 4. METHOD 2: INVERSE-PROBABILITY WEIGHTING (IPW) ===" === 4. METHOD 2: INVERSE-PROBABILITY WEIGHTING (IPW) === . . * teffects ipw: probit treatment model, POMs, ATE, ATT . teffects ipw (bweight) (mbsmoke mmarried mage fbaby medu, probit), pomeans no > log Treatment-effects estimation Number of obs = 4,642 Estimator : inverse-probability weights Outcome model : weighted mean Treatment model: probit ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- POmeans | mbsmoke | Nonsmoker | 3402.429 9.576032 355.31 0.000 3383.66 3421.197 Smoker | 3171.523 22.37227 141.76 0.000 3127.674 3215.372 ------------------------------------------------------------------------------ . teffects ipw (bweight) (mbsmoke mmarried mage fbaby medu, probit), ate nol > og Treatment-effects estimation Number of obs = 4,642 Estimator : inverse-probability weights Outcome model : weighted mean Treatment model: probit ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATE | mbsmoke | (Smoker | vs | Nonsmoker) | -230.906 24.30987 -9.50 0.000 -278.5525 -183.2595 -------------+---------------------------------------------------------------- POmean | mbsmoke | Nonsmoker | 3402.429 9.576032 355.31 0.000 3383.66 3421.197 ------------------------------------------------------------------------------ . estimates store te_ipw . teffects ipw (bweight) (mbsmoke mmarried mage fbaby medu, probit), atet nol > og Treatment-effects estimation Number of obs = 4,642 Estimator : inverse-probability weights Outcome model : weighted mean Treatment model: probit ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATET | mbsmoke | (Smoker | vs | Nonsmoker) | -219.6338 23.38456 -9.39 0.000 -265.4667 -173.8009 -------------+---------------------------------------------------------------- POmean | mbsmoke | Nonsmoker | 3357.294 13.76189 243.96 0.000 3330.321 3384.266 ------------------------------------------------------------------------------ . estimates store te_ipw_att . . * Manual recreation: estimate propensity scores, build IPW weights, weighted > regression . di _newline "--- Manual recreation of IPW ---" --- Manual recreation of IPW --- . logistic mbsmoke mmarried mage fbaby medu, nolog Logistic regression Number of obs = 4,642 LR chi2(4) = 346.31 Prob > chi2 = 0.0000 Log likelihood = -2057.5951 Pseudo R2 = 0.0776 ------------------------------------------------------------------------------ mbsmoke | Odds ratio Std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- mmarried | .354782 .0319308 -11.51 0.000 .2974078 .4232244 mage | .9963441 .0083578 -0.44 0.662 .980097 1.01286 fbaby | .6209276 .0541327 -5.47 0.000 .5233991 .7366293 medu | .8805241 .0147636 -7.59 0.000 .8520582 .909941 _cons | 2.786206 .6821068 4.19 0.000 1.724352 4.50195 ------------------------------------------------------------------------------ Note: _cons estimates baseline odds. . predict ps, p . generate ipw_w = 1/ps if mbsmoke==1 (3,778 missing values generated) . replace ipw_w = 1/(1-ps) if mbsmoke==0 (3,778 real changes made) . list bweight mbsmoke ps ipw_w in 40/49 +-------------------------------------------+ | bweight mbsmoke ps ipw_w | |-------------------------------------------| 40. | 2790 Nonsmoker .0596716 1.063458 | 41. | 4082 Nonsmoker .1067052 1.119451 | 42. | 3089 Nonsmoker .0933409 1.10295 | 43. | 2580 Smoker .3121328 3.203764 | 44. | 3714 Nonsmoker .1277395 1.146447 | |-------------------------------------------| 45. | 4196 Nonsmoker .1019658 1.113543 | 46. | 2523 Nonsmoker .3417279 1.519129 | 47. | 2693 Smoker .3557771 2.810749 | 48. | 3204 Nonsmoker .0676847 1.072599 | 49. | 3572 Smoker .2595506 3.852813 | +-------------------------------------------+ . . regress bweight mbsmoke [aweight=ipw_w] (sum of wgt is 9,154.62682104111) Source | SS df MS Number of obs = 4,642 -------------+---------------------------------- F(1, 4640) = 192.15 Model | 62512147.2 1 62512147.2 Prob > F = 0.0000 Residual | 1.5095e+09 4,640 325332.601 R-squared = 0.0398 -------------+---------------------------------- Adj R-squared = 0.0396 Total | 1.5721e+09 4,641 338732.044 Root MSE = 570.38 ------------------------------------------------------------------------------ bweight | Coefficient Std. err. t P>|t| [95% conf. interval] -------------+---------------------------------------------------------------- mbsmoke | -232.1279 16.74591 -13.86 0.000 -264.9578 -199.2979 _cons | 3402.621 11.73618 289.93 0.000 3379.613 3425.63 ------------------------------------------------------------------------------ . di "Manual IPW estimate (coefficient on mbsmoke): " %7.2f _b[mbsmoke] " grams > " Manual IPW estimate (coefficient on mbsmoke): -232.13 grams . . * Figure 2: propensity-score distribution by treatment status . twoway /// > (histogram ps if mbsmoke==0, fcolor("106 155 204%50") lcolor("106 155 204 > ") width(0.025)) /// > (histogram ps if mbsmoke==1, fcolor("217 119 87%50") lcolor("217 119 87" > ) width(0.025)) /// > , /// > title("Estimated Propensity Scores", size(medium)) /// > subtitle("Probability of smoking, by actual smoking status", size(small)) > /// > xtitle("Estimated propensity score, e(X)") ytitle("Density") /// > legend(order(1 "Non-smokers (D=0)" 2 "Smokers (D=1)") position(6) rows(1) > ) /// > graphregion(color(white)) plotregion(color(white)) /// > note("Where the two distributions overlap, we can credibly compare smoker > s and non-smokers.", size(vsmall)) . graph export "stata_matching_propensity_distribution.png", replace width(2400 > ) file stata_matching_propensity_distribution.png written in PNG format . . *--------------------------------------------------- . * 5. Method 3: Inverse-Probability-Weighted Regression Adjustment (IPWRA) . *--------------------------------------------------- . . di _newline(2) "=== 5. METHOD 3: IPWRA ===" === 5. METHOD 3: IPWRA === . . * IPWRA: combines outcome model + treatment model (doubly robust) . teffects ipwra (bweight mmarried mage prenatal1 fbaby) /// > (mbsmoke mmarried mage fbaby medu, probit), pomeans aequations > nolog Treatment-effects estimation Number of obs = 4,642 Estimator : IPW regression adjustment Outcome model : linear Treatment model: probit ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- POmeans | mbsmoke | Nonsmoker | 3402.699 9.570442 355.54 0.000 3383.941 3421.456 Smoker | 3170.826 23.29268 136.13 0.000 3125.173 3216.479 -------------+---------------------------------------------------------------- OME0 | mmarried | 156.7215 26.62711 5.89 0.000 104.5334 208.9097 mage | 3.119249 2.158588 1.45 0.148 -1.111506 7.350005 prenatal1 | 66.72463 28.58028 2.33 0.020 10.70831 122.7409 fbaby | -71.64215 20.38959 -3.51 0.000 -111.605 -31.67929 _cons | 3188.271 55.72367 57.22 0.000 3079.055 3297.488 -------------+---------------------------------------------------------------- OME1 | mmarried | 137.657 39.85045 3.45 0.001 59.55152 215.7624 mage | -6.380565 5.420729 -1.18 0.239 -17.005 4.243868 prenatal1 | 28.52348 41.70547 0.68 0.494 -53.21774 110.2647 fbaby | 51.35695 51.04662 1.01 0.314 -48.6926 151.4065 _cons | 3198.271 138.9068 23.02 0.000 2926.019 3470.524 -------------+---------------------------------------------------------------- TME1 | mmarried | -.5848065 .0532985 -10.97 0.000 -.6892696 -.4803433 mage | -.0023797 .0048267 -0.49 0.622 -.0118399 .0070804 fbaby | -.2610501 .0492315 -5.30 0.000 -.3575421 -.164558 medu | -.0784464 .0095783 -8.19 0.000 -.0972195 -.0596734 _cons | .6177995 .1380763 4.47 0.000 .3471751 .888424 ------------------------------------------------------------------------------ . teffects ipwra (bweight mmarried mage prenatal1 fbaby) /// > (mbsmoke mmarried mage fbaby medu, probit), ate nolog Treatment-effects estimation Number of obs = 4,642 Estimator : IPW regression adjustment Outcome model : linear Treatment model: probit ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATE | mbsmoke | (Smoker | vs | Nonsmoker) | -231.8723 25.1541 -9.22 0.000 -281.1735 -182.5712 -------------+---------------------------------------------------------------- POmean | mbsmoke | Nonsmoker | 3402.699 9.570442 355.54 0.000 3383.941 3421.456 ------------------------------------------------------------------------------ . estimates store te_ipwra . teffects ipwra (bweight mmarried mage prenatal1 fbaby) /// > (mbsmoke mmarried mage fbaby medu, probit), atet nolog Treatment-effects estimation Number of obs = 4,642 Estimator : IPW regression adjustment Outcome model : linear Treatment model: probit ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATET | mbsmoke | (Smoker | vs | Nonsmoker) | -220.6476 23.37268 -9.44 0.000 -266.4572 -174.838 -------------+---------------------------------------------------------------- POmean | mbsmoke | Nonsmoker | 3358.307 13.78516 243.62 0.000 3331.289 3385.326 ------------------------------------------------------------------------------ . estimates store te_ipwra_att . . *--------------------------------------------------- . * 6. Method 4: Augmented Inverse-Probability Weighting (AIPW) . *--------------------------------------------------- . . di _newline(2) "=== 6. METHOD 4: AIPW ===" === 6. METHOD 4: AIPW === . . * AIPW: efficient and doubly robust; in Stata, ATE only. . teffects aipw (bweight mmarried mage prenatal1 fbaby) /// > (mbsmoke mmarried mage fbaby medu, probit), pomeans aequations > nolog Treatment-effects estimation Number of obs = 4,642 Estimator : augmented IPW Outcome model : linear by ML Treatment model: probit ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- POmeans | mbsmoke | Nonsmoker | 3402.688 9.57038 355.54 0.000 3383.93 3421.445 Smoker | 3170.212 22.9462 138.16 0.000 3125.238 3215.186 -------------+---------------------------------------------------------------- OME0 | mmarried | 160.9513 26.6162 6.05 0.000 108.7845 213.1181 mage | 2.546828 2.084324 1.22 0.222 -1.538373 6.632028 prenatal1 | 64.40859 27.52699 2.34 0.019 10.45669 118.3605 fbaby | -71.3286 19.64701 -3.63 0.000 -109.836 -32.82117 _cons | 3202.746 54.01082 59.30 0.000 3096.886 3308.605 -------------+---------------------------------------------------------------- OME1 | mmarried | 133.6617 40.86443 3.27 0.001 53.5689 213.7545 mage | -7.370881 4.21817 -1.75 0.081 -15.63834 .8965804 prenatal1 | 25.11133 40.37541 0.62 0.534 -54.02302 104.2457 fbaby | 41.43991 39.70712 1.04 0.297 -36.38461 119.2644 _cons | 3227.169 104.4059 30.91 0.000 3022.537 3431.801 -------------+---------------------------------------------------------------- TME1 | mmarried | -.5848065 .0532985 -10.97 0.000 -.6892696 -.4803433 mage | -.0023797 .0048267 -0.49 0.622 -.0118399 .0070804 fbaby | -.2610501 .0492315 -5.30 0.000 -.3575421 -.164558 medu | -.0784464 .0095783 -8.19 0.000 -.0972195 -.0596734 _cons | .6177995 .1380763 4.47 0.000 .3471751 .888424 ------------------------------------------------------------------------------ . teffects aipw (bweight mmarried mage prenatal1 fbaby) /// > (mbsmoke mmarried mage fbaby medu, probit), ate nolog Treatment-effects estimation Number of obs = 4,642 Estimator : augmented IPW Outcome model : linear by ML Treatment model: probit ------------------------------------------------------------------------------ | Robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATE | mbsmoke | (Smoker | vs | Nonsmoker) | -232.4759 24.83406 -9.36 0.000 -281.1497 -183.802 -------------+---------------------------------------------------------------- POmean | mbsmoke | Nonsmoker | 3402.688 9.57038 355.54 0.000 3383.93 3421.445 ------------------------------------------------------------------------------ . estimates store te_aipw . . *--------------------------------------------------- . * 7. Method 5: Nearest-Neighbor Matching (NNM) . *--------------------------------------------------- . . di _newline(2) "=== 7. METHOD 5: NEAREST-NEIGHBOR MATCHING (NNM) ===" === 7. METHOD 5: NEAREST-NEIGHBOR MATCHING (NNM) === . . * Basic NNM (one nearest neighbor on Mahalanobis distance) . teffects nnmatch (bweight mmarried mage fage medu prenatal1) (mbsmoke), nolog Treatment-effects estimation Number of obs = 4,642 Estimator : nearest-neighbor matching Matches: requested = 1 Outcome model : matching min = 1 Distance metric: Mahalanobis max = 16 ------------------------------------------------------------------------------ | AI robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATE | mbsmoke | (Smoker | vs | Nonsmoker) | -210.5435 29.32969 -7.18 0.000 -268.0286 -153.0584 ------------------------------------------------------------------------------ . . * NNM with exact matching on discrete covariates and bias adjustment on conti > nuous covariates . teffects nnmatch (bweight mmarried mage fage medu prenatal1) (mbsmoke), /// > ematch(mmarried prenatal1) biasadj(mage fage medu) nolog Treatment-effects estimation Number of obs = 4,642 Estimator : nearest-neighbor matching Matches: requested = 1 Outcome model : matching min = 1 Distance metric: Mahalanobis max = 16 ------------------------------------------------------------------------------ | AI robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATE | mbsmoke | (Smoker | vs | Nonsmoker) | -210.0558 29.32803 -7.16 0.000 -267.5377 -152.5739 ------------------------------------------------------------------------------ . estimates store te_nnmatch . teffects nnmatch (bweight mmarried mage fage medu prenatal1) (mbsmoke), /// > ematch(mmarried prenatal1) biasadj(mage fage medu) atet nolog Treatment-effects estimation Number of obs = 4,642 Estimator : nearest-neighbor matching Matches: requested = 1 Outcome model : matching min = 1 Distance metric: Mahalanobis max = 16 ------------------------------------------------------------------------------ | AI robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATET | mbsmoke | (Smoker | vs | Nonsmoker) | -238.5204 30.41661 -7.84 0.000 -298.1359 -178.905 ------------------------------------------------------------------------------ . estimates store te_nnmatch_att . . *--------------------------------------------------- . * 8. Method 6: Propensity-Score Matching (PSM) . *--------------------------------------------------- . . di _newline(2) "=== 8. METHOD 6: PROPENSITY-SCORE MATCHING (PSM) ===" === 8. METHOD 6: PROPENSITY-SCORE MATCHING (PSM) === . . * Visualize the matching idea on a small subsample (Figure 3) . * Using the propensity score we already computed (ps) . preserve . keep if _n < 100 (4,543 observations deleted) . #delimit ; delimiter now ; . twoway > (scatter mbsmoke ps if mbsmoke==0, > msize(medium) msymbol(circle_hollow) mcolor("106 155 204")) > (scatter mbsmoke ps if mbsmoke==1, > msize(medium) msymbol(circle_hollow) mcolor("217 119 87")) > (pcarrowi 0.92 0.09 0.08 0.09, > recast(pcbarrow) lcolor("20 20 20") lwidth(medium) > mcolor("20 20 20") msize(medium) barbsize(medium)) > , > title("How Propensity-Score Matching Works", size(medium)) > subtitle("Match a smoker with the most similar non-smoker(s)", size(s > mall)) > text(0.50 0.20 > "Match each smoker with the nearest non-smoker(s) {&rarr}" > "in propensity-score space", > color("20 20 20") just(left) placement(e) size(small)) > ytitle("") > yscale(range(-0.1 1.1)) > ylabel(0 "Non-smoker" 1 "Smoker", angle(horizontal)) > xtitle("Estimated propensity score, e(X)") xlabel(0(0.1)1) > legend(off) > graphregion(color(white)) plotregion(color(white)) > ; . #delimit cr delimiter now cr . graph export "stata_matching_psm_logic.png", replace width(2400) file stata_matching_psm_logic.png written in PNG format . restore . . * PSM estimation . teffects psmatch (bweight) (mbsmoke mmarried mage fage medu prenatal1), nolog Treatment-effects estimation Number of obs = 4,642 Estimator : propensity-score matching Matches: requested = 1 Outcome model : matching min = 1 Treatment model: logit max = 16 ------------------------------------------------------------------------------ | AI robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATE | mbsmoke | (Smoker | vs | Nonsmoker) | -229.4492 25.88746 -8.86 0.000 -280.1877 -178.7107 ------------------------------------------------------------------------------ . estimates store te_psmatch . teffects psmatch (bweight) (mbsmoke mmarried mage fage medu prenatal1), atet > nolog Treatment-effects estimation Number of obs = 4,642 Estimator : propensity-score matching Matches: requested = 1 Outcome model : matching min = 1 Treatment model: logit max = 16 ------------------------------------------------------------------------------ | AI robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATET | mbsmoke | (Smoker | vs | Nonsmoker) | -224.5927 30.55147 -7.35 0.000 -284.4725 -164.7129 ------------------------------------------------------------------------------ . estimates store te_psmatch_att . . * Overlap diagnostic (Figure 4) . teffects psmatch (bweight) (mbsmoke mmarried mage fage medu prenatal1), nolog Treatment-effects estimation Number of obs = 4,642 Estimator : propensity-score matching Matches: requested = 1 Outcome model : matching min = 1 Treatment model: logit max = 16 ------------------------------------------------------------------------------ | AI robust bweight | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- ATE | mbsmoke | (Smoker | vs | Nonsmoker) | -229.4492 25.88746 -8.86 0.000 -280.1877 -178.7107 ------------------------------------------------------------------------------ . teffects overlap, /// > ptlevel(0) /// > title("Overlap of Propensity Scores", size(medium)) /// > subtitle("Required for valid causal comparison", size(small)) /// > legend(order(1 "Non-smokers (D=0)" 2 "Smokers (D=1)") position(6) rows(1) > ) /// > graphregion(color(white)) plotregion(color(white)) (refitting the model using the generate() option) . graph export "stata_matching_overlap.png", replace width(2400) file stata_matching_overlap.png written in PNG format . . *--------------------------------------------------- . * 9. Comparison of all six methods . *--------------------------------------------------- . . di _newline(2) "=== 9. COMPARISON OF ALL SIX METHODS ===" === 9. COMPARISON OF ALL SIX METHODS === . . * ATE comparison table for the six teffects estimators . di _newline "--- ATE estimates from teffects estimators ---" --- ATE estimates from teffects estimators --- . estimates table te_ra te_ipw te_ipwra te_aipw te_nnmatch te_psmatch /// > , b(%9.2f) se(%7.2f) stats(N) varwidth(16) modelwidth(11) -------------------------------------------------------------------------- Variable | te_ra te_ipw te_ipwra te_aipw -----------------+-------------------------------------------------------- ATE | mbsmoke | (Smoker | vs | Nonsmoker) | -239.64 -230.91 -231.87 -232.48 | 23.82 24.31 25.15 24.83 -----------------+-------------------------------------------------------- POmean | mbsmoke | Nonsmoker | 3403.24 3402.43 3402.70 3402.69 | 9.53 9.58 9.57 9.57 -----------------+-------------------------------------------------------- OME0 | mmarried | 160.95 156.72 160.95 | 26.62 26.63 26.62 mage | 2.55 3.12 2.55 | 2.08 2.16 2.08 prenatal1 | 64.41 66.72 64.41 | 27.53 28.58 27.53 fbaby | -71.33 -71.64 -71.33 | 19.65 20.39 19.65 _cons | 3202.75 3188.27 3202.75 | 54.01 55.72 54.01 -----------------+-------------------------------------------------------- OME1 | mmarried | 133.66 137.66 133.66 | 40.86 39.85 40.86 mage | -7.37 -6.38 -7.37 | 4.22 5.42 4.22 prenatal1 | 25.11 28.52 25.11 | 40.38 41.71 40.38 fbaby | 41.44 51.36 41.44 | 39.71 51.05 39.71 _cons | 3227.17 3198.27 3227.17 | 104.41 138.91 104.41 -----------------+-------------------------------------------------------- TME1 | mmarried | -0.58 -0.58 -0.58 | 0.05 0.05 0.05 mage | -0.00 -0.00 -0.00 | 0.00 0.00 0.00 fbaby | -0.26 -0.26 -0.26 | 0.05 0.05 0.05 medu | -0.08 -0.08 -0.08 | 0.01 0.01 0.01 _cons | 0.62 0.62 0.62 | 0.14 0.14 0.14 -----------------+-------------------------------------------------------- Statistics | N | 4642 4642 4642 4642 -------------------------------------------------------------------------- Legend: b/se ---------------------------------------------- Variable | te_nnmatch te_psmatch -----------------+---------------------------- ATE | mbsmoke | (Smoker | vs | Nonsmoker) | -210.06 -229.45 | 29.33 25.89 -----------------+---------------------------- POmean | mbsmoke | Nonsmoker | | -----------------+---------------------------- OME0 | mmarried | | mage | | prenatal1 | | fbaby | | _cons | | -----------------+---------------------------- OME1 | mmarried | | mage | | prenatal1 | | fbaby | | _cons | | -----------------+---------------------------- TME1 | mmarried | | mage | | fbaby | | medu | | _cons | | -----------------+---------------------------- Statistics | N | 4642 4642 ---------------------------------------------- Legend: b/se . . * Naive baseline shown separately (different coefficient name in regress) . di _newline "--- Naive (unadjusted) baseline ---" --- Naive (unadjusted) baseline --- . estimates table te_naive, b(%9.2f) se(%7.2f) stats(N) varwidth(16) modelwidth > (11) -------------------------------- Variable | te_naive -----------------+-------------- mbsmoke | -275.25 | 21.22 _cons | 3412.91 | 9.29 -----------------+-------------- N | 4642 -------------------------------- Legend: b/se . . * ATT comparison table (note: AIPW does not provide ATT in Stata) . di _newline "--- ATT estimates (five methods) ---" --- ATT estimates (five methods) --- . estimates table te_ra_att te_ipw_att te_ipwra_att te_nnmatch_att te_psmatch_a > tt /// > , b(%9.2f) se(%7.2f) stats(N) varwidth(16) modelwidth(11) -------------------------------------------------------------------------- Variable | te_ra_att te_ipw_att te_ipwra_~t te_nnmatc~t -----------------+-------------------------------------------------------- ATET | mbsmoke | (Smoker | vs | Nonsmoker) | -223.30 -219.63 -220.65 -238.52 | 22.74 23.38 23.37 30.42 -----------------+-------------------------------------------------------- POmean | mbsmoke | Nonsmoker | 3360.96 3357.29 3358.31 | 12.76 13.76 13.79 -----------------+-------------------------------------------------------- OME0 | mmarried | 160.95 139.32 | 26.62 29.17 mage | 2.55 5.28 | 2.08 2.92 prenatal1 | 64.41 73.34 | 27.53 37.02 fbaby | -71.33 -73.97 | 19.65 30.93 _cons | 3202.75 3136.41 | 54.01 72.56 -----------------+-------------------------------------------------------- OME1 | mmarried | 133.66 133.66 | 40.86 40.86 mage | -7.37 -7.37 | 4.22 4.22 prenatal1 | 25.11 25.11 | 40.38 40.38 fbaby | 41.44 41.44 | 39.71 39.71 _cons | 3227.17 3227.17 | 104.41 104.41 -----------------+-------------------------------------------------------- TME1 | mmarried | -0.58 -0.58 | 0.05 0.05 mage | -0.00 -0.00 | 0.00 0.00 fbaby | -0.26 -0.26 | 0.05 0.05 medu | -0.08 -0.08 | 0.01 0.01 _cons | 0.62 0.62 | 0.14 0.14 -----------------+-------------------------------------------------------- Statistics | N | 4642 4642 4642 4642 -------------------------------------------------------------------------- Legend: b/se -------------------------------- Variable | te_psmatc~t -----------------+-------------- ATET | mbsmoke | (Smoker | vs | Nonsmoker) | -224.59 | 30.55 -----------------+-------------- POmean | mbsmoke | Nonsmoker | | -----------------+-------------- OME0 | mmarried | | mage | | prenatal1 | | fbaby | | _cons | | -----------------+-------------- OME1 | mmarried | | mage | | prenatal1 | | fbaby | | _cons | | -----------------+-------------- TME1 | mmarried | | mage | | fbaby | | medu | | _cons | | -----------------+-------------- Statistics | N | 4642 -------------------------------- Legend: b/se . . * Figure 5: forest plot of ATE estimates . * coefplot's rename + keep doesn't span regress (mbsmoke) and teffects . * (r1vs0.mbsmoke) cleanly, so we collect ATE and 95% CI manually via postfile > . . * Collect b, ll, ul, and a row index manually from each stored estimate . tempfile fp_data . capture postclose fp . postfile fp str20 method double(b ll ul) byte row using `fp_data', replace (file /var/folders/f_/4_1h2nwn2w91_4qmnp8p1snm0000gn/T//St60458.000003 not found) . . estimates restore te_naive (results te_naive are active now) . post fp ("0. Naive") (_b[mbsmoke]) (_b[mbsmoke] - 1.96*_se[mbsmo > ke]) (_b[mbsmoke] + 1.96*_se[mbsmoke]) (7) . . estimates restore te_ra (results te_ra are active now) . post fp ("1. RA") (_b[r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] - 1.96*_se > [r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] + 1.96*_se[r1vs0.mbsmoke]) (6) . . estimates restore te_ipw (results te_ipw are active now) . post fp ("2. IPW") (_b[r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] - 1.96*_se > [r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] + 1.96*_se[r1vs0.mbsmoke]) (5) . . estimates restore te_ipwra (results te_ipwra are active now) . post fp ("3. IPWRA") (_b[r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] - 1.96*_se > [r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] + 1.96*_se[r1vs0.mbsmoke]) (4) . . estimates restore te_aipw (results te_aipw are active now) . post fp ("4. AIPW") (_b[r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] - 1.96*_se > [r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] + 1.96*_se[r1vs0.mbsmoke]) (3) . . estimates restore te_nnmatch (results te_nnmatch are active now) . post fp ("5. NNM") (_b[r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] - 1.96*_se > [r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] + 1.96*_se[r1vs0.mbsmoke]) (2) . . estimates restore te_psmatch (results te_psmatch are active now) . post fp ("6. PSM") (_b[r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] - 1.96*_se > [r1vs0.mbsmoke]) (_b[r1vs0.mbsmoke] + 1.96*_se[r1vs0.mbsmoke]) (1) . . postclose fp . . preserve . use `fp_data', clear . list method b ll ul, sep(0) noobs +-------------------------------------------------+ | method b ll ul | |-------------------------------------------------| | 0. Naive -275.25187 -316.83329 -233.67046 | | 1. RA -239.63921 -286.33429 -192.94413 | | 2. IPW -230.90597 -278.55333 -183.25862 | | 3. IPWRA -231.87233 -281.17438 -182.57029 | | 4. AIPW -232.47587 -281.15064 -183.8011 | | 5. NNM -210.05578 -267.53873 -152.57284 | | 6. PSM -229.44922 -280.18865 -178.7098 | +-------------------------------------------------+ . export delimited using "ate_estimates.csv", replace file ate_estimates.csv saved . . * Build the forest plot . twoway /// > (rcap ll ul row, horizontal lcolor("106 155 204") lwidth(medthick)) / > // > (scatter row b, msymbol(D) mcolor("217 119 87") msize(large)) /// > , /// > xline(0, lcolor("20 20 20") lpattern(dash)) /// > ylabel(7 "0. Naive" 6 "1. RA" 5 "2. IPW" 4 "3. IPWRA" 3 "4. AIPW" /// > 2 "5. NNM" 1 "6. PSM", angle(horizontal) labsize(small) nogrid > ) /// > ytitle("") /// > xtitle("ATE on birth weight (grams)") /// > title("Estimated Effect of Maternal Smoking on Birth Weight", size(me > dium)) /// > subtitle("Six estimators + naive baseline, with 95% confidence interv > als", size(small)) /// > legend(off) /// > graphregion(color(white)) plotregion(color(white)) /// > note("Negative values: smoking lowers birth weight. The naive estimat > e is biased; the six estimators move the answer closer to the truth.", size(v > small)) . graph export "stata_matching_forest_plot.png", replace width(2400) file stata_matching_forest_plot.png written in PNG format . restore . . *--------------------------------------------------- . * 10. End of analysis . *--------------------------------------------------- . . di _newline(2) . di "============================================================" ============================================================ . di " Analysis complete." Analysis complete. . di "" . di " Six teffects estimators compared on cattaneo2.dta:" Six teffects estimators compared on cattaneo2.dta: . di " Naive : " %7.2f naive_b " grams" Naive : -275.25 grams . di " RA, IPW, IPWRA, AIPW, NNM, PSM: see comparison table" RA, IPW, IPWRA, AIPW, NNM, PSM: see comparison table . di "" . di " Outputs:" Outputs: . di " - analysis.log" - analysis.log . di " - stata_matching_density_bweight.png" - stata_matching_density_bweight.png . di " - stata_matching_propensity_distribution.png" - stata_matching_propensity_distribution.png . di " - stata_matching_psm_logic.png" - stata_matching_psm_logic.png . di " - stata_matching_overlap.png" - stata_matching_overlap.png . di " - stata_matching_forest_plot.png" - stata_matching_forest_plot.png . di "" . di " Script completed successfully" Script completed successfully . di "============================================================" ============================================================ . . log close name: log: /Users/carlos/GitHub/starter-academic-v501/content/post/stata_matc > hing/analysis.log log type: text closed on: 29 Apr 2026, 12:22:54 -------------------------------------------------------------------------------