Fig. 1From: Using marginal standardisation to estimate relative risk without dichotomising continuous outcomesSimulations results for bias, coverage probability and standard error for both \( \hat{\beta_1} \) and \( \hat{\mathrm{RR}} \). Panel (a) and (b) plot the simulation results for \( \hat{\beta_1} \) and panel (c) and (d) plot the simulation results for \( \hat{\mathrm{RR}} \), when errors are normally (or logistically) distributed with mean (or location) 0 and standard deviation (or scale) 1, and α1 = 0,−0.15, and −0.3. Dashed lines are 0, 0.95 and 1 for Bias, Coverage probability and Ratio: Mean/Empirical standard error (SE) respectively, which correspond to no bias, 95% coverage probability and mean and empirical SEs are the same. Normal and logistic linear models mean linear model with the error terms assumed to have normal and logistic distribution respectivelyBack to article page