20210802_170604 study03 canada and usa results

```r # screener > dt1[, .N, keyby = .(nationality, screener_pass)] nationality screener_pass N 1: americans -0.5 151 2: americans 0.5 1102 3: canadian -0.5 145 4: canadian 0.5 1088 ``` ```r # Dem/Rep per study > dt1[, .N, keyby = .(nationality, demrep)] nationality demrep N 1: americans 1 573 2: americans 2 529 3: canadian 1 891 4: canadian 2 197 ``` # Treatments 4 and 13 ![[Pasted image 20210711122924.png|600]] # Correlations ![[Pasted image 20210802172014.png|500]] ![[Pasted image 20210802171926.png|1200]] # Discernment Treatment 4 increased discernment in both Dem/Rep, but treatment 13 reduced Rep's overall sharing. ![[fake_real_demrep 3.png|800]] ![[fake_real_demrep_dist 2.png|800]] ## Discernment models No main effect of condition (control vs 2 treatments combined) ```r > dt1[, summaryh(lm(discern ~ condition)), by = .(nationality)] nationality term results 1: canadian (Intercept) b = 0.85, SE = 0.03, t(1086) = 26.04, p < .001, r = 0.62 2: canadian condition b = 0.08, SE = 0.07, t(1086) = 1.15, p = .252, r = 0.04 # 3: americans (Intercept) b = 0.59, SE = 0.04, t(1100) = 16.23, p < .001, r = 0.44 4: americans condition b = 0.07, SE = 0.07, t(1100) = 0.97, p = .330, r = 0.03 # # combined > dt1[, summaryh(lm(discern ~ condition))] term results 1: (Intercept) b = 0.72, SE = 0.02, t(2188) = 29.23, p < .001, r = 0.53 2: condition b = 0.07, SE = 0.05, t(2188) = 1.45, p = .147, r = 0.03 # ``` Weak effect of treatment 4 (vs control [intercept]) in combined results. ```r > dt1[, summaryh(lm(discern ~ treatment)), by = .(nationality)] nationality term results 1: canadian (Intercept) b = 0.82, SE = 0.05, t(1085) = 15.31, p < .001, r = 0.42 2: canadian treatment4 b = 0.12, SE = 0.08, t(1085) = 1.56, p = .118, r = 0.05 # 3: canadian treatment13 b = 0.03, SE = 0.08, t(1085) = 0.42, p = .674, r = 0.01 4: americans (Intercept) b = 0.56, SE = 0.06, t(1099) = 9.35, p < .001, r = 0.27 5: americans treatment4 b = 0.14, SE = 0.08, t(1099) = 1.73, p = .085, r = 0.05 # 6: americans treatment13 b = −0.005, SE = 0.08, t(1099) = −0.06, p = .953, r = 0 # combined > dt1[, summaryh(lm(discern ~ treatment))] term results 1: (Intercept) b = 0.69, SE = 0.04, t(2187) = 17.07, p < .001, r = 0.34 2: treatment4 b = 0.13, SE = 0.06, t(2187) = 2.27, p = .024, r = 0.05 # 3: treatment13 b = 0.01, SE = 0.06, t(2187) = 0.23, p = .817, r = 0 ``` Canadians more discerning than Americans, but likely because Canadian sample has more Democrats (see next chunk). ```r > summaryh(lm(discern ~ condition * nationalityEC, dt1)) term results 1: (Intercept) b = 0.72, SE = 0.02, t(2186) = 29.46, p < .001, r = 0.53 2: condition b = 0.07, SE = 0.05, t(2186) = 1.49, p = .136, r = 0.03 3: nationalityEC b = 0.26, SE = 0.05, t(2186) = 5.32, p < .001, r = 0.11 4: condition:nationalityEC b = 0.004, SE = 0.10, t(2186) = 0.04, p = .968, r = 0 ``` Added `demrepEC` (party) to above model. Canadians no longer more discerning. ```r > summaryh(lm(discern ~ condition * nationalityEC + demrepEC, dt1)) term results 1: (Intercept) b = 0.57, SE = 0.02, t(2185) = 23.79, p < .001, r = 0.45 2: condition b = 0.04, SE = 0.05, t(2185) = 0.97, p = .332, r = 0.02 3: nationalityEC b = −0.01, SE = 0.05, t(2185) = −0.20, p = .842, r = 0 # 4: demrepEC b = −0.91, SE = 0.05, t(2185) = −18.89, p < .001, r = −0.38 # 5: condition:nationalityEC b = −0.007, SE = 0.09, t(2185) = −0.07, p = .941, r = 0 ``` No interaction between treatment and party. ```r > dt1[, summaryh(lm(discern ~ treatmentEC * demrepEC)), by = .(nationality)] nationality term results 1: canadian (Intercept) b = 0.69, SE = 0.05, t(715) = 13.97, p < .001, r = 0.46 2: canadian treatmentEC b = −0.06, SE = 0.10, t(715) = −0.65, p = .515, r = −0.02 3: canadian demrepEC b = −0.60, SE = 0.10, t(715) = −6.05, p < .001, r = −0.22 4: canadian treatmentEC:demrepEC b = 0.005, SE = 0.20, t(715) = 0.02, p = .982, r = 0 5: americans (Intercept) b = 0.60, SE = 0.04, t(730) = 16.43, p < .001, r = 0.52 6: americans treatmentEC b = −0.14, SE = 0.07, t(730) = −1.98, p = .048, r = −0.07 7: americans demrepEC b = −1.03, SE = 0.07, t(730) = −14.10, p < .001, r = −0.46 8: americans treatmentEC:demrepEC b = 0.23, SE = 0.15, t(730) = 1.58, p = .114, r = 0.06 ``` # Treatment outcomes Dems prefer 4, Reps prefer 13. ![[treatment_dvs.png|800]] # Discernment: determine party with `feeling_therm_right` (support Trump/Toole) If `feeling_therm_right >= 50, Republican, else Democrat`. ```r # no. of participants > dt1[!is.na(feeling_therm_right2), .N, keyby = .(nationality, feeling_therm_right2)] nationality feeling_therm_right2 N 1: americans -0.5 657 2: americans 0.5 445 3: canadian -0.5 685 4: canadian 0.5 402 > dt1[!is.na(feeling_therm_right2), .N, keyby = .(nationality, treatment, feeling_therm_right2)] nationality treatment feeling_therm_right2 N 1: americans 0 -0.5 215 2: americans 0 0.5 153 3: americans 4 -0.5 228 4: americans 4 0.5 145 5: americans 13 -0.5 214 6: americans 13 0.5 147 7: canadian 0 -0.5 221 8: canadian 0 0.5 148 9: canadian 4 -0.5 238 10: canadian 4 0.5 120 11: canadian 13 -0.5 226 12: canadian 13 0.5 134 ``` Significant interaction in American dataset. Treatment 13 is more effective for those who support Trump. ```r > dt1[, summaryh(lm(discern ~ treatmentEC * feeling_therm_right2)), by = .(nationality)] # significant nationality term results 1: canadian (Intercept) b = 0.83, SE = 0.04, t(714) = 21.38, p < .001, r = 0.62 2: canadian treatmentEC b = −0.08, SE = 0.08, t(714) = −0.97, p = .331, r = −0.04 3: canadian feeling_therm_right2 b = −0.41, SE = 0.08, t(714) = −5.22, p < .001, r = −0.19 4: canadian treatmentEC:feeling_therm_right2 b = −0.03, SE = 0.16, t(714) = −0.20, p = .843, r = −0.01 5: americans (Intercept) b = 0.52, SE = 0.04, t(730) = 13.98, p < .001, r = 0.46 6: americans treatmentEC b = −0.10, SE = 0.07, t(730) = −1.29, p = .198, r = −0.05 7: americans feeling_therm_right2 b = −1.01, SE = 0.07, t(730) = −13.54, p < .001, r = −0.45 8: americans treatmentEC:feeling_therm_right2 b = 0.34, SE = 0.15, t(730) = 2.30, p = .022, r = 0.09 # ``` ![[fake_real__feeling_support_right2.png|800]] ## Similar-ish results with `party` ```r > dt1[, summaryh(lm(discern ~ treatmentEC * partyEC)), by = .(nationality)] # significant nationality term results 1: canadian (Intercept) b = 0.74, SE = 0.08, t(463) = 9.29, p < .001, r = 0.40 2: canadian treatmentEC b = −0.28, SE = 0.16, t(463) = −1.76, p = .079, r = −0.08 3: canadian partyEC b = −0.61, SE = 0.16, t(463) = −3.84, p < .001, r = −0.18 4: canadian treatmentEC:partyEC b = −0.31, SE = 0.32, t(463) = −0.96, p = .337, r = −0.04 5: americans (Intercept) b = 0.60, SE = 0.04, t(701) = 16.10, p < .001, r = 0.52 6: americans treatmentEC b = −0.13, SE = 0.07, t(701) = −1.79, p = .073, r = −0.07 7: americans partyEC b = −1.07, SE = 0.07, t(701) = −14.43, p < .001, r = −0.48 8: americans treatmentEC:partyEC b = 0.25, SE = 0.15, t(701) = 1.68, p = .094, r = 0.06 # # no. of participants > dt1[!is.na(partyEC), .N, keyby = .(nationality, partyEC)] nationality partyEC N 1: americans -0.5 545 2: americans 0.5 509 3: canadian -0.5 606 4: canadian 0.5 88 ```