241013_210648 compare dave vs causal forest results - struct racism data

# models ## dv: t1 ```r # dave model (two elastic nets) > feols(t1 ~ IR1treated_1 * predictEffect + t0 , d0) NOTE: 330 observations removed because of NA values (LHS: 312, RHS: 285). OLS estimation, Dep. Var.: t1 Observations: 1,007 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) -1.748152 1.693844 -1.032062 3.0229e-01 IR1treated_1 0.059175 1.602383 0.036929 9.7055e-01 predictEffect 0.625977 0.130460 4.798235 1.8439e-06 *** t0 1.019428 0.020857 48.877983 < 2.2e-16 *** IR1treated_1:predictEffect 0.985361 0.166309 5.924865 4.2934e-09 *** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 RMSE: 17.8 Adj. R2: 0.746196 # dave model (hause implementation using two separate regression forests to make predictions separately for control vs treatment groups) > feols(t1 ~ IR1treated_1 * tau_diff + t0 , d0) NOTE: 312 observations removed because of NA values (LHS: 312, RHS: 265). OLS estimation, Dep. Var.: t1 Observations: 1,025 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 14.128791 1.749525 8.075786 1.8796e-15 *** IR1treated_1 -0.592912 2.106134 -0.281517 7.7837e-01 tau_diff -0.608772 0.137149 -4.438756 1.0033e-05 *** t0 0.867677 0.017996 48.213945 < 2.2e-16 *** IR1treated_1:tau_diff 0.856271 0.199006 4.302739 1.8492e-05 *** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 RMSE: 18.9 Adj. R2: 0.712463 # caual forests approach (one causal forest) > feols(t1 ~ IR1treated_1 * tau + t0 , d0) NOTE: 312 observations removed because of NA values (LHS: 312, RHS: 265). OLS estimation, Dep. Var.: t1 Observations: 1,025 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 17.157834 2.684406 6.39167 2.4921e-10 *** IR1treated_1 -6.082724 3.492921 -1.74144 8.1907e-02 . tau -1.285057 0.345783 -3.71636 2.1309e-04 *** t0 0.871451 0.018035 48.32047 < 2.2e-16 *** IR1treated_1:tau 1.948314 0.491208 3.96638 7.8073e-05 *** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 RMSE: 19.0 Adj. R2: 0.711031 ``` ## dv: t2 ```r # dave two elastic nets > feols(t2 ~ IR1treated_1 * predictEffect + t0, d0[!is.na(t1)]) NOTE: 386 observations removed because of NA values (LHS: 374, RHS: 18). OLS estimation, Dep. Var.: t2 Observations: 639 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 6.005669 2.673178 2.24664 0.025006 * IR1treated_1 -2.829944 2.533274 -1.11711 0.264371 predictEffect 0.315472 0.207002 1.52401 0.128006 t0 0.800892 0.032803 24.41510 < 2.2e-16 *** IR1treated_1:predictEffect 0.618410 0.258989 2.38779 0.017242 * --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 RMSE: 22.0 Adj. R2: 0.544112 # two regression forests > feols(t2 ~ IR1treated_1 * tau_diff + t0, d0[!is.na(t1)]) NOTE: 374 observations removed because of NA values (LHS: 374). OLS estimation, Dep. Var.: t2 Observations: 651 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 21.586472 2.434420 8.86719 < 2.2e-16 *** IR1treated_1 5.898261 2.483837 2.37466 1.7856e-02 * tau_diff -1.051628 0.218839 -4.80548 1.9215e-06 *** t0 0.639937 0.027994 22.85960 < 2.2e-16 *** IR1treated_1:tau_diff -0.845833 0.302112 -2.79973 5.2673e-03 ** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 RMSE: 20.9 Adj. R2: 0.591462 # one causal forest > feols(t2 ~ IR1treated_1 * tau + t0, d0[!is.na(t1)]) NOTE: 374 observations removed because of NA values (LHS: 374). OLS estimation, Dep. Var.: t2 Observations: 651 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 12.127887 3.760503 3.225070 1.3229e-03 ** IR1treated_1 19.391869 4.582853 4.231397 2.6587e-05 *** tau -0.018830 0.501387 -0.037555 9.7005e-01 t0 0.711800 0.027316 26.058156 < 2.2e-16 *** IR1treated_1:tau -2.956928 0.695064 -4.254179 2.4086e-05 *** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 RMSE: 21.7 Adj. R2: 0.559638 ``` ## dv: t2Imp ```r # dave two elastic nets > feols(t2Imp ~ IR1treated_1 * predictEffect + t0, d0[!is.na(t1)]) NOTE: 18 observations removed because of NA values (RHS: 18). OLS estimation, Dep. Var.: t2Imp Observations: 1,007 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 3.858529 1.725796 2.23580 0.0255847 * IR1treated_1 -1.665025 1.632609 -1.01986 0.3080433 predictEffect 0.190180 0.132921 1.43078 0.1528060 t0 0.877893 0.021250 41.31260 < 2.2e-16 *** IR1treated_1:predictEffect 0.437769 0.169447 2.58352 0.0099204 ** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 RMSE: 18.2 Adj. R2: 0.700042 # two regression forests > feols(t2Imp ~ IR1treated_1 * tau_diff + t0, d0[!is.na(t1)]) OLS estimation, Dep. Var.: t2Imp Observations: 1,025 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 8.545252 1.659746 5.14853 3.1492e-07 *** IR1treated_1 6.843452 1.998055 3.42506 6.3915e-04 *** tau_diff -0.142742 0.130111 -1.09707 2.7287e-01 t0 0.829994 0.017073 48.61473 < 2.2e-16 *** IR1treated_1:tau_diff -0.653723 0.188794 -3.46263 5.5712e-04 *** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 RMSE: 18.0 Adj. R2: 0.707922 # one causal forest > feols(t2Imp ~ IR1treated_1 * tau + t0, d0[!is.na(t1)]) OLS estimation, Dep. Var.: t2Imp Observations: 1,025 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 3.911594 2.547144 1.53568 1.2493e-01 IR1treated_1 16.097547 3.314318 4.85697 1.3787e-06 *** tau 0.479655 0.328102 1.46191 1.4407e-01 t0 0.833744 0.017113 48.72092 < 2.2e-16 *** IR1treated_1:tau -2.222077 0.466091 -4.76748 2.1364e-06 *** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 RMSE: 18.0 Adj. R2: 0.706354 ``` ![[cor 1.png]]