Comparison of Parametric and Non-Parametric Methods for Determining Injury Risk

This paper contains a review of methods for deriving risk curves from biomechanical data obtained from impact experiments on human surrogates. It covers many of the problems and pitfalls of obtaining realistic human risk curves from impact experiments. The strength and weakness of both parametric and non-parametric methods are evaluated. The limitations of standard analysis of censored impact test data are presented. Methods are given for determining risk curves from both doubly censored data and data obtained from impacts to body regions in which there are more than one mechanism of injury. A detailed set of examples is presented in which different experimental data are analyzed using the Consistent Threshold method and the logistic approach. Finally risk curves for published data are presented for the femur, head, thorax, and neck.