# Discernment ~ treatment, agree, inform, like, share ![[discern_tmtresponse_treatment.png]] ![[fakesm_tmtresponse_treatment.png]] ![[realsm_tmtresponse_treatment.png]] # Regression results (from machine learning with cross-validation) ## Study 1 ### Target: `discern` - model with 30 (out of 34 possible) features explained most variance (21%) in unseen data - top treatments: 10, 3, 8 - best features: sei, tmt_agree, tmt_like, tmt_share, feeling_therm_trump - interactions: sei treatment ```r > summaryh(lm(discern ~ tm_inform + tm_share + tm_like + tm_agree, dt1_1)) term results 1: (Intercept) b = 0.34, SE = 0.09, t(2284) = 3.70, p < .001, r = 0.08 2: tm_inform b = −0.16, SE = 0.02, t(2284) = −7.35, p < .001, r = −0.15 3: tm_share b = 0.15, SE = 0.09, t(2284) = 1.76, p = .078, r = 0.04 4: tm_like b = 0.41, SE = 0.08, t(2284) = 4.88, p < .001, r = 0.10 5: tm_agree b = 0.20, SE = 0.02, t(2284) = 9.58, p < .001, r = 0.20 ``` ### Target: `tmt_agree` - top treatments: 10, 3, 8 - best features: sei, feeling therm_trump, fb_share_politic, social_conserv - interactions: sei treatment, ### Target: `tmt_inform` - top treatments: 10, 4, 3, 2, 8, 6, 7 - best feature: demrep, education, sei - interactions: demrep treatment ```r > summaryh(aov(tm_inform ~ as.factor(treatment) * demrep, dt1_1)) term results 1: as.factor(treatment) F(12, 2269) = 19.22, p < .001, r = 0.30 2: demrep F(1, 2269) = 16.93, p < .001, r = 0.09 3: as.factor(treatment):demrep F(12, 2269) = 1.71, p = .059, r = 0.09 ``` ### Target: `tmt_like` - top treatments: 9, 8, 10, 4, 6, 3, 12 - best features: fb_share_politic, sei, demrep, education - interactions: ### Target: `tmt_share` - top treatments: 8, 4, 3, 5, 6, 5, 10 - best features: fb_share_politic, demrep, education, sei, economic_conserv - interactions: ## Study 2 ### Target: `discern` ### Target: `tmt_agree` ### Target: `tmt_inform` ### Target: `tmt_like` ### Target: `tmt_share`