The utility of the epsilon-skew-normal distribution in the context of receiver operating characteristic (ROC) curve fitting
Terry Mashtare, PhD University at Buffalo
In this talk, we extend the well-known parametric normal model via implementation of the epsilon-skew-normal (ESN) distribution developed by Mudholkar and Hutson (2000). We derive the equation for the receiver operating characteristic (ROC) curve assuming a parametric ESN model and examine the behavior of the maximum likelihood estimates for estimating the model parameters. We then summarize the results of a simulation study to examine the asymptotic properties of the maximum likelihood estimates in the parametric ESN model and compare with the maximum likelihood estimates in the normal model. We also summarize the results of a simulation study comparing the two parametric models to the nonparametric ROC model. We then illustrate the maximum likelihood estimation of the parametric ESN model using data involving skeletal measurements in 507 physically active individuals.