A non-parametric method for hazard rate estimation in acute myocardial infarction patients: Kernel smoothing approach

Kernel smoothing method is a non-parametric or graphical method for statistical estimation. In the present study was used a kernel smoothing method for finding the death hazard rates of patients with acute myocardial infarction. By employing non-parametric regression methods, the curve estimation, may have some complexity. In this article, four indices of Epanechnikov, Biquadratic, Triquadratic and Rectangle kernels were used under local and k-nearest neighbor's bandwidth. For comparing the models, were employed mean integrated squared error. To illustrate in the study, was used the dataset of acute myocardial infraction patients in Bushehr port, in the south of Iran. To obtain proper bandwidth, was used generalized cross-validation method. Corresponding to a low bandwidth value, the curve is unreadable and the regression curve is so roughly. In the event of increasing bandwidth value, the distribution has more readable and smooth. In this study, estimate of death hazard rate for the patients based on Epanechnikov kernel under local bandwidth was 1.011×10-11, which had the lowest mean square error compared to k-nearest neighbors bandwidth. We obtained the death hazard rate in 10 and 30 months after the first acute myocardial infraction using Epanechnikov kernelas were 0.0031 and 0.0012, respectively. The Epanechnikov kernel for obtaining death hazard rate of patients with acute myocardial infraction has minimum mean integrated squared error compared to the other kernels. In addition, the mortality hazard rate of acute myocardial infraction in the study was low.