Skip to main content
Fig. 3 | BMC Medical Research Methodology

Fig. 3

From: Time dependent hazard ratio estimation using instrumental variables without conditioning on an omitted covariate

Fig. 3

Simulation Results for Estimating a Log-Linear Time Dependent Hazard Ratio: The top 3 panes represent estimation of the intercept parameter of the log-linear model for (i) strong positive confounding, (ii) no confounding and (iii) strong negative confounding, while the bottom three panes represent results for estimation of the slope parameter. Each panel overlays scatterplots of the median bias (estimated hazard ratio minus true coefficient) versus the true coefficient, our IV based estimator in black, and the Cox MPLE in red. Each point represents the median of over 1000 estimates (from that many simulated datasets)

Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us