- [[240903_162316 effect size|effect size harris vote minus trump vote]] Let $t_i p_i$ be trump vote and $(1-t_i)p_i$ be harris vote. - $\text{diff}_{pre} = ((1-t_i)p_i)_{pre} - (t_i p_i)_{pre}$ - `harris_trump_vote_1` in code - $\text{diff}_{post} = ((1-t_i)p_i)_{post} - (t_i p_i)_{post}$ - `harris_trump_vote_2` in code - $\text{diff harris trump vote} = \text{diff}_{post} - \text{diff}_{pre}$ - `diff_harris_trump_vote` in code below ```r # separate models for each condition > d1[, summ(lm(diff_harris_trump_vote ~ 1)), by = .(condition)] condition term result <char> <char> <char> 1: persuadeHarris (Intercept) b = 0.03 (0.01) [0.02, 0.04] p < .001 2: persuadeTrump (Intercept) b = -0.03 (0.01) [-0.04, -0.02] p < .001 # separate models for each condition and topic > d1[, summ(lm(diff_harris_trump_vote ~ 1)), by = .(condition, topic)] condition topic term result <char> <char> <char> <char> 1: persuadeHarris policy (Intercept) b = 0.04 (0.01) [0.02, 0.05] p < .001 2: persuadeHarris personality (Intercept) b = 0.02 (0.01) [0.00, 0.04] p = .047 3: persuadeTrump policy (Intercept) b = -0.03 (0.01) [-0.05, -0.02] p < .001 4: persuadeTrump personality (Intercept) b = -0.03 (0.01) [-0.04, -0.01] p < .001 # extra stuff (other model specifications) # diff ~ pre * condition > d1[, summ(lm(diff_harris_trump_vote ~ condition * topicZ * harris_trump_vote_1Z))] term result <char> <char> 1: (Intercept) b = 0.03 (0.01) [0.02, 0.04] p < .001 2: conditionpersuadeTrump b = -0.06 (0.01) [-0.07, -0.04] p < .001 3: topicZ b = 0.01 (0.01) [0.00, 0.02] p = .108 4: harris_trump_vote_1Z b = -0.01 (0.01) [-0.03, 0.00] p = .010 5: conditionpersuadeTrump × topicZ b = -0.01 (0.01) [-0.03, 0.00] p = .114 6: conditionpersuadeTrump × harris_trump_vote_1Z b = 0.01 (0.01) [0.00, 0.03] p = .145 7: topicZ × harris_trump_vote_1Z b = 0.00 (0.01) [-0.01, 0.01] p = .772 8: conditionpersuadeTrump × topicZ × harris_trump_vote_1Z b = 0.00 (0.01) [-0.02, 0.01] p = .646 # diff ~ pre * condition + 0 (remove intercept) > d1[, summ(lm(diff_harris_trump_vote ~ 0 + condition * topicZ * harris_trump_vote_1Z))] term result <char> <char> 1: conditionpersuadeHarris b = 0.03 (0.01) [0.02, 0.04] p < .001 2: conditionpersuadeTrump b = -0.03 (0.01) [-0.04, -0.02] p < .001 3: topicZ b = 0.01 (0.01) [0.00, 0.02] p = .108 4: harris_trump_vote_1Z b = -0.01 (0.01) [-0.03, 0.00] p = .010 5: conditionpersuadeTrump × topicZ b = -0.01 (0.01) [-0.03, 0.00] p = .114 6: conditionpersuadeTrump × harris_trump_vote_1Z b = 0.01 (0.01) [0.00, 0.03] p = .145 7: topicZ × harris_trump_vote_1Z b = 0.00 (0.01) [-0.01, 0.01] p = .772 8: conditionpersuadeTrump × topicZ × harris_trump_vote_1Z b = 0.00 (0.01) [-0.02, 0.01] p = .646 # post ~ pre * condition > summ(lm(harris_trump_vote_2 ~ condition * topicZ * harris_trump_vote_1Z, d1)) term result <char> <char> 1: (Intercept) b = 0.16 (0.01) [0.15, 0.17] p < .001 2: conditionpersuadeTrump b = -0.06 (0.01) [-0.07, -0.04] p < .001 3: topicZ b = 0.01 (0.01) [0.00, 0.02] p = .108 4: harris_trump_vote_1Z b = 0.78 (0.01) [0.77, 0.79] p < .001 5: conditionpersuadeTrump × topicZ b = -0.01 (0.01) [-0.03, 0.00] p = .114 6: conditionpersuadeTrump × harris_trump_vote_1Z b = 0.01 (0.01) [0.00, 0.03] p = .145 7: topicZ × harris_trump_vote_1Z b = 0.00 (0.01) [-0.01, 0.01] p = .772 8: conditionpersuadeTrump × topicZ × harris_trump_vote_1Z b = 0.00 (0.01) [-0.02, 0.01] p = .646 ``` # previous calculations Let $t_i p_i$ (trump vote) and $(1-t_i)p_i$ (harris vote). - $\text{diff trump vote} = (t_i p_i)_{post} - (t_i p_i)_{pre}$ - $\text{diff harris vote} = ((1-t_i)p_i)_{post} - ((1-t_i)p_i)_{pre}$ ```r > summ(feols(diff_harris_vote ~ condition, d1)) term result <char> <char> 1: (Intercept) b = 0.02 (0.00) [0.02, 0.03] p < .001 2: conditionpersuadeTrump b = -0.03 (0.00) [-0.04, -0.02] p < .001 > summ(feols(diff_trump_vote ~ condition, d1)) term result <char> <char> 1: (Intercept) b = -0.01 (0.00) [-0.01, 0.00] p = .020 2: conditionpersuadeTrump b = 0.03 (0.00) [0.02, 0.04] p < .001 > summ(feols(diff_harris_vote ~ condition * topicZ, d1)) term result <char> <char> 1: (Intercept) b = 0.02 (0.00) [0.02, 0.03] p < .001 2: conditionpersuadeTrump b = -0.03 (0.00) [-0.04, -0.02] p < .001 3: topicZ b = 0.00 (0.00) [0.00, 0.01] p = .360 4: conditionpersuadeTrump × topicZ b = -0.01 (0.00) [-0.01, 0.00] p = .262 > summ(feols(diff_trump_vote ~ condition * topicZ, d1)) term result <char> <char> 1: (Intercept) b = -0.01 (0.00) [-0.01, 0.00] p = .021 2: conditionpersuadeTrump b = 0.03 (0.00) [0.02, 0.04] p < .001 3: topicZ b = -0.01 (0.00) [-0.01, 0.00] p = .065 4: conditionpersuadeTrump × topicZ b = 0.01 (0.00) [0.00, 0.02] p = .106 # maybe the models below aren't as kosher since initial lean was used to calculate the outcome itself > summ(feols(diff_harris_vote ~ condition * topicZ * lean_bidentrump_1Z, d1)) term result <char> <char> 1: (Intercept) b = 0.02 (0.00) [0.02, 0.03] p < .001 2: conditionpersuadeTrump b = -0.03 (0.00) [-0.04, -0.02] p < .001 3: topicZ b = 0.00 (0.00) [0.00, 0.01] p = .357 4: lean_bidentrump_1Z b = 0.00 (0.00) [0.00, 0.01] p = .552 5: conditionpersuadeTrump × topicZ b = -0.01 (0.00) [-0.01, 0.00] p = .263 6: conditionpersuadeTrump × lean_bidentrump_1Z b = 0.00 (0.00) [0.00, 0.01] p = .379 7: topicZ × lean_bidentrump_1Z b = 0.00 (0.00) [-0.01, 0.01] p = .918 8: conditionpersuadeTrump × topicZ × lean_bidentrump_1Z b = 0.00 (0.00) [-0.01, 0.01] p = .689 > summ(feols(diff_trump_vote ~ condition * topicZ * lean_bidentrump_1Z, d1)) term result <char> <char> 1: (Intercept) b = -0.01 (0.00) [-0.01, 0.00] p = .023 2: conditionpersuadeTrump b = 0.03 (0.00) [0.02, 0.04] p < .001 3: topicZ b = -0.01 (0.00) [-0.01, 0.00] p = .063 4: lean_bidentrump_1Z b = -0.01 (0.00) [-0.01, 0.00] p = .010 5: conditionpersuadeTrump × topicZ b = 0.01 (0.00) [0.00, 0.02] p = .106 6: conditionpersuadeTrump × lean_bidentrump_1Z b = 0.02 (0.00) [0.01, 0.02] p = .001 7: topicZ × lean_bidentrump_1Z b = 0.00 (0.00) [-0.01, 0.01] p = .951 8: conditionpersuadeTrump × topicZ × lean_bidentrump_1Z b = 0.00 (0.00) [-0.01, 0.01] p = .697 ```