Fig. 4
From: Using marginal standardisation to estimate relative risk without dichotomising continuous outcomes
Effect of intervention on risk of hyperglycaemia with respect to different degree of severity. Panel (a) and (b) plot \( \hat{\beta_1} \) s and \( \hat{\mathrm{RR}} \) s with their corresponding 95% confidence intervals (i.e., vertical lines) for linear, probit and logit models respectively. * indicates a P-value less than 0.05 without adjustment for multiple testing. + indicates a P-value less than 0.05 after adjustment for multiple testing