A nonparametric data based univariate density function estimate (Q1084798)

A method for assigning a value to the smoothing parameter in the kernel estimate of a univariate density function is derived. The method uses observations of the random variable whose probability density function is to be estimated, and it is not sensitive to outliers in the data. Simulation studies were performed to compare the performance of the method with others that have been proposed. These studies indicate that the method produces density estimators which have smaller mean integrated squared errors for a wide variety of types of density functions.