Figure 11
Logistic regression models for intention to download the Corona-Warn-App (Waves 3 and 4). Dependent variable: intention to download the app in the future (yes/no). Coefficients: measures from the survey (e.g., combined score for trust in app security; combined score for conspiracy beliefs; see Supplementary Table B10). Horizontal point ranges indicate point estimates and 95% confidence intervals for each predictor. Education was dummy coded with the reference level “medium,” yielding two coefficients: low (vs. medium) and high (vs. medium) education. Following30, we standardized all continuous variables by two standard deviations (SD) and mean centered the binary gender variable. This way a 2-SD change in a continuous predictor variable is approximately equivalent to a change of category in a roughly balanced binary predictor variable (e.g., gender). In a logistic regression model, a slope reflects the relative change in log odds (while keeping all other predictors at their average values). Supplementary Table A7 summarizes the regression results for both models. Supplementary Figure A9 provides an alternative arrangement of the same results, where the respective model for each wave is shown in a separate panel.
