Compute propensity scores $p(condition = treatment | covariates)$ with logistic regression model: `condition ~ 10 covariates` (see below for covariates). # those who completed both days N = 275 ```r condition N 1: accuracy 183 2: interest 92 ``` Probability of being assigned to treatment condition is higher in the treatment (accuracy) than control (interest) condition. ```r # logistic regression predicting treatment assignment with covariates m0 <- glm(conditiond ~ age + gender + region + country + domain + role + tenuredays + aot + covid_concern + conspiracy, d00, family = "binomial") d00$propensity_glm <- predict(m0, type = 'response') # t-test > summaryh(t.test(propensity_glm ~ condition, d00)) results 1: t(174) = 8.60, p < .001, r = 0.55 # multilevel model with counterbalance as clustering variable > summaryh(m1 <- lmer(propensity_glm ~ condition + (1 | counterbalance), d00)) term results 1: (Intercept) b = 0.70, SE = 0.04, t(6) = 16.21, p < .001, r = 0.99 2: conditioninterest b = −0.18, SE = 0.06, t(7) = −2.90, p = .024, r = 0.75 ``` There's quite a bit of overlap in the propensity score distributions for treatment and control groups, so we should be good (but maybe match on propensity scores)? ![[_temp.png]] Propensity scores for different conditions (8 conditions for counterbalancing). ![[_temp 2.png]] # extra check: those who completed only day 1 or both days N = 360 ```r condition N 1: accuracy 245 2: interest 115 ``` ```r # logistic regression predicting treatment assignment with covariates m0 <- glm(conditiond ~ age + gender + region + country + domain + role + tenuredays + aot + covid_concern + conspiracy, d00, family = "binomial") d00$propensity_glm <- predict(m0, type = 'response') # t-test > summaryh(t.test(propensity_glm ~ condition, d00)) results 1: t(232) = 8.62, p < .001, r = 0.49 # multilevel model with counterbalance as clustering variable > summaryh(m1 <- lmer(propensity_glm ~ condition + (1 | counterbalance), d00)) term results 1: (Intercept) b = 0.70, SE = 0.04, t(6) = 18.42, p < .001, r = 0.99 2: conditioninterest b = −0.14, SE = 0.05, t(7) = −2.52, p = .042, r = 0.70 ``` ![[_temp 1.png]] ![[_temp 3.png]]