Fig. 3: Equivalence test results for studies 1, 2a, and 2b.
From: Warning people that they are being microtargeted fails to eliminate persuasive advantage
Results for the equivalence tests of Study 1 (left), Study 2a (center), and Study 2b (right). The tests are calculated based on the linear mixed-effects models described in the Methods section, and include the effects of ad type (popup and non-targeted), the centered matching score, and the fully-crossed interaction between the two. The point ranges for each variable correspond to the model estimate and its 90% narrow confidence intervals. The light blue area designates the region of practical equivalence (ROPE), which in these two cases corresponds to [-0.126, 0.126]. Practical equivalence is accepted (in red) when the narrow confidence intervals are completely within the ROPE, regardless of statistical significance. Practical equivalence is instead rejected (in green) when the coefficient is statistically significant, whether the narrow confidence intervals (i.e., 1 − 2*α) include or exclude the ROPE boundaries, as long as the narrow confidence intervals are not fully covered by the ROPE. Finally, the decision to accept or reject practical equivalence is undecided (in yellow) when effects are not statistically significant and the narrow confidence intervals overlap the ROPE.
