``` > d2[, n_distinct(user_id), condition] condition V1 1: -0.5 136 2: 0.5 112 ``` # correlations between DVs ```r > cors topic r 1: climate 0.8818548 2: election 0.7929602 3: johnny 0.6816853 ``` ![[s20220523_095637.png]] # manipulation check: how many % climate cards seen ![[s20220523_095719.png]] # main results No effect of treatment on climate importance, newsworthiness. Looks like ceiling effects? ```r > m1_imp <- lm(importance ~ condition, d3[topic == "climate"]) > summaryh(m1_imp) term results 1: (Intercept) b = 5.04, SE = 0.08, t(246) = 60.51, p < .001, r = 0.97 2: condition b = 0.02, SE = 0.17, t(246) = 0.14, p = .885, r = 0.01 BF = 0.14, 0.051 > m1_newsw <- lm(newsworthiness ~ condition, d3[topic == "climate"]) > summaryh(m1_newsw) term results 1: (Intercept) b = 4.81, SE = 0.09, t(246) = 56.03, p < .001, r = 0.96 2: condition b = −0.005, SE = 0.17, t(246) = −0.03, p = .976, r = 0 BF = 0.1397504, 0.051 > summaryh(m1_combine) term results 1: (Intercept) b = 4.92, SE = 0.08, t(246) = 60.03, p < .001, r = 0.97 2: condition b = 0.009, SE = 0.16, t(246) = 0.06, p = .954, r = 0 BF = 0.1399083, 0.051 ``` But potential transfer effects? Treatment group thinks election is less important/newsworthy? ```r > summaryh(lm(importance ~ condition, d3[topic == "election"])) term results 1: (Intercept) b = 4.82, SE = 0.07, t(246) = 72.54, p < .001, r = 0.98 2: condition b = −0.22, SE = 0.13, t(246) = −1.65, p = .100, r = −0.10 > summaryh(lm(newsworthiness ~ condition, d3[topic == "election"])) term results 1: (Intercept) b = 4.74, SE = 0.07, t(246) = 70.69, p < .001, r = 0.98 2: condition b = −0.29, SE = 0.13, t(246) = −2.19, p = .030, r = −0.14 > summaryh(lm(rating_combined ~ condition, d3[topic == "election"])) term results 1: (Intercept) b = 4.78, SE = 0.06, t(246) = 75.69, p < .001, r = 0.98 2: condition b = −0.26, SE = 0.13, t(246) = −2.03, p = .043, r = −0.13 ``` ![[main_effect.png]] ## perform analysis on subset of subjects who reported seeing "correct" no. of headlines ```r d3a <- d3[(condition == -0.5 & climate_perc < 20) | (condition == 0.5 & climate_perc > 33)] condition V1 1: -0.5 82 2: 0.5 95 ``` ![[manipulation_check_subset 1.png]] **Reduced** importance, newsworthiness, problem, human activity, concern? Potential backfire? ![[main_effect_subset.png]] ## feed engagement associated with outcomes x-axis: proportion of climate headlines shared/liked/bookmarked (only in the treatment group) ![[s20220523_105854.png]] ![[s20220523_110049.png]] ![[s20220523_110118.png]] ![[s20220523_110145.png]] ![[s20220523_110222.png]] ![[s20220523_110250.png]] ![[s20220523_110310.png]] ![[s20220523_110430.png]] ![[s20220523_110636.png]]