```r # participants condition N <char> <int> 1: catdog 541 2: personality 532 3: policy 519 # vote vote_bidentrump N <fctr> <int> 1: harris 909 2: trump 378 3: other 305 # affiliation pol demrep N <int> <num> <int> 1: 1 1.0 224 2: 2 1.0 300 3: 3 1.0 356 4: 4 1.5 428 5: 5 2.0 175 6: 6 2.0 73 7: 7 2.0 36 # words typed condition text1 text2 text_total <char> <num> <num> <num> 1: catdog 37.15342 38.55453 75.70795 2: personality 43.26692 36.21617 79.48308 3: policy 45.16378 35.42967 80.59345 ``` # results - modeling: condition: catdog (intercept), personality, policy - suffix `Z`: zscored variable ## main models - `t2 ~ condition * t1Z` across all participants ```r > feols(lean_bidentrump_2 ~ condition * lean_bidentrump_1Z, data = d0) OLS estimation, Dep. Var.: lean_bidentrump_2 Observations: 1,592 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 33.001504 0.314425 104.958198 < 2.2e-16 *** conditionpersonality -0.017791 0.446738 -0.039825 0.968238 conditionpolicy 0.365570 0.449559 0.813176 0.416239 lean_bidentrump_1Z 38.674578 0.311238 124.260532 < 2.2e-16 *** conditionpersonality:lean_bidentrump_1Z 0.730377 0.440347 1.658640 0.097386 . # personality treatment made trump voters lean more trump? conditionpolicy:lean_bidentrump_1Z 0.619888 0.454926 1.362614 0.173198 > feols(vote_chance_2 ~ condition * vote_chance_1Z, data = d0) OLS estimation, Dep. Var.: vote_chance_2 Observations: 1,592 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 81.738571 0.277147 294.928999 < 2.2e-16 *** conditionpersonality 0.764023 0.393483 1.941692 0.05235131 . conditionpolicy 0.083144 0.396092 0.209911 0.83376416 vote_chance_1Z 31.206672 0.270805 115.236540 < 2.2e-16 *** conditionpersonality:vote_chance_1Z -1.282606 0.387742 -3.307888 0.00096106 *** # less likely to vote after personality treatment for those who were more likely to vote initially? conditionpolicy:vote_chance_1Z -0.029279 0.398698 -0.073436 0.94146819 # nothing for feeling thermometers > feols(therm_dem_2 ~ condition * therm_dem_1Z, data = d0) OLS estimation, Dep. Var.: therm_dem_2 Observations: 1,592 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 59.451735 0.432018 137.614010 < 2.2e-16 *** conditionpersonality 0.202726 0.614272 0.330026 0.74142 conditionpolicy 0.068248 0.618380 0.110365 0.91213 therm_dem_1Z 26.960998 0.436699 61.738162 < 2.2e-16 *** conditionpersonality:therm_dem_1Z 0.112780 0.612020 0.184274 0.85382 conditionpolicy:therm_dem_1Z -0.426124 0.624501 -0.682343 0.49512 > feols(therm_repub_2 ~ condition * therm_repub_1Z, data = d0) OLS estimation, Dep. Var.: therm_repub_2 Observations: 1,592 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 36.389974 0.432912 84.058670 < 2.2e-16 *** conditionpersonality 0.236108 0.614806 0.384036 0.70100 conditionpolicy 0.859151 0.618639 1.388776 0.16510 therm_repub_1Z 28.289386 0.428861 65.964053 < 2.2e-16 *** conditionpersonality:therm_repub_1Z -0.846054 0.612769 -1.380706 0.16756 conditionpolicy:therm_repub_1Z 0.551438 0.618451 0.891643 0.37272 ``` ## difference scores for all four variables - columns: different variables/outcome variables (difference scores: post minus pre) - rows: pre-treatment candidate preference (harris, trump, other voters) ![[1722093061.png]] ## cubic interactions ### lean biden/harris vs trump Same interactions as before, but nothing significant ```r # outcome: lean term estimate std.error statistic p.value <chr> <dbl> <dbl> <dbl> <dbl> 1 (Intercept) 0.962 0.444 2.17 0.0303 2 conditionpersonality -0.685 0.639 -1.07 0.284 3 conditionpolicy 0.0357 0.634 0.0563 0.955 4 lean_bidentrump_1 0.732 0.0868 8.43 0 5 I(lean_bidentrump_1^2) 0.00599 0.00244 2.45 0.0142 6 I(lean_bidentrump_1^3) -0.00004 0.00002 -2.20 0.0282 7 conditionpersonality:lean_bidentrump_1 0.0991 0.125 0.795 0.427 8 conditionpolicy:lean_bidentrump_1 -0.138 0.124 -1.11 0.266 9 conditionpersonality:I(lean_bidentrump_1^2) -0.00261 0.00348 -0.751 0.453 10 conditionpolicy:I(lean_bidentrump_1^2) 0.0048 0.00353 1.36 0.174 11 conditionpersonality:I(lean_bidentrump_1^3) 0.00002 0.00002 0.788 0.431 12 conditionpolicy:I(lean_bidentrump_1^3) -0.00003 0.00002 -1.39 0.166 ``` ![[1722092466.png]] ![[1722264540.png]] ### vote likelihood (interesting results?) Very significant cubic interaction (see catdog vs personality conditions) ```r # vote chance/likelihood term estimate std.error statistic p.value <chr> <dbl> <dbl> <dbl> <dbl> 1 (Intercept) 0.897 1.00 0.896 0.370 2 conditionpersonality 0.791 1.47 0.540 0.589 3 conditionpolicy -0.519 1.50 -0.345 0.730 4 vote_chance_1 0.784 0.107 7.33 0 5 I(vote_chance_1^2) 0.0051 0.00249 2.04 0.0411 6 I(vote_chance_1^3) -0.00003 0.00002 -2.00 0.0461 7 conditionpersonality:vote_chance_1 0.500 0.159 3.15 0.00167 8 conditionpolicy:vote_chance_1 0.0312 0.159 0.196 0.844 9 conditionpersonality:I(vote_chance_1^2) -0.0116 0.00371 -3.14 0.00173 10 conditionpolicy:I(vote_chance_1^2) 0.00016 0.00377 0.0428 0.966 11 conditionpersonality:I(vote_chance_1^3) 0.00007 0.00002 2.89 0.00396 # more cubic for personality treatment 12 conditionpolicy:I(vote_chance_1^3) 0 0.00002 -0.184 0.854 # interact condition also with policy_importance (new question we added), not cubic - just linear terms > feols(vote_chance_2 ~ condition * (vote_chance_1Z + scale(policy_imp_1)), data = d0) NOTE: 568 observations removed because of NA values (RHS: 568). OLS estimation, Dep. Var.: vote_chance_2 Observations: 1,024 Standard-errors: IID Estimate Std. Error t value Pr(>|t|) (Intercept) 81.719131 0.338404 241.484239 < 2.2e-16 *** conditionpersonality 0.627146 0.484079 1.295544 0.1954271 conditionpolicy 0.038081 0.480915 0.079185 0.9369014 vote_chance_1Z 31.084483 0.382712 81.221509 < 2.2e-16 *** scale(policy_imp_1) 0.055739 0.354198 0.157368 0.8749861 conditionpersonality:vote_chance_1Z -0.455067 0.523469 -0.869329 0.3848727 conditionpolicy:vote_chance_1Z 0.215491 0.534380 0.403254 0.6868466 conditionpersonality:scale(policy_imp_1) -1.498153 0.508399 -2.946807 0.0032841 ** # personality treatment reduced voting likelihood among those who care more about policy conditionpolicy:scale(policy_imp_1) -0.277280 0.515268 -0.538128 0.5906067 ``` ![[1722092536.png]] ![[1722264582.png]] ### thermometer democrat ```r # A tibble: 12 × 5 term estimate std.error statistic p.value <chr> <dbl> <dbl> <dbl> <dbl> 1 (Intercept) 3.45 1.74 1.98 0.0476 2 conditionpersonality 0.256 2.35 0.109 0.914 3 conditionpolicy -1.41 2.51 -0.562 0.574 4 therm_dem_1 0.793 0.144 5.51 0 5 I(therm_dem_1^2) 0.00549 0.00327 1.68 0.0935 6 I(therm_dem_1^3) -0.00004 0.00002 -1.90 0.0571 7 conditionpersonality:therm_dem_1 -0.0146 0.198 -0.0738 0.941 8 conditionpolicy:therm_dem_1 0.257 0.205 1.25 0.211 9 conditionpersonality:I(therm_dem_1^2) 0.00032 0.00456 0.0693 0.945 10 conditionpolicy:I(therm_dem_1^2) -0.00623 0.00464 -1.34 0.179 11 conditionpersonality:I(therm_dem_1^3) 0 0.00003 -0.0567 0.955 12 conditionpolicy:I(therm_dem_1^3) 0.00004 0.00003 1.31 0.189 ``` ![[1722093400.png]] ![[1722264644.png]] ### thermometer republican ```r # A tibble: 12 × 5 term estimate std.error statistic p.value <chr> <dbl> <dbl> <dbl> <dbl> 1 (Intercept) 1.17 0.955 1.23 0.219 2 conditionpersonality 3.18 1.44 2.22 0.0268 3 conditionpolicy 0.237 1.40 0.169 0.865 4 therm_repub_1 0.874 0.109 8.00 0 5 I(therm_repub_1^2) 0.00345 0.00294 1.18 0.240 6 I(therm_repub_1^3) -0.00003 0.00002 -1.25 0.212 7 conditionpersonality:therm_repub_1 -0.249 0.158 -1.57 0.116 8 conditionpolicy:therm_repub_1 0.00302 0.156 0.0194 0.985 9 conditionpersonality:I(therm_repub_1^2) 0.00472 0.00422 1.12 0.263 10 conditionpolicy:I(therm_repub_1^2) 0.00053 0.00413 0.129 0.898 11 conditionpersonality:I(therm_repub_1^3) -0.00003 0.00003 -0.889 0.374 12 conditionpolicy:I(therm_repub_1^3) 0 0.00003 -0.144 0.886 ``` ![[1722093462.png]] ![[1722264682.png]]