Optimal histogram interval for non-stationary time-series distribution function density estimation
Abstract:
The properties of distributions of the distances between two empirical distribution function densities for non-stationary time-series are investigated. The optimal choice of histogram interval is suggested on the basis of self-agreement accuracy of empirical probability. The estimation is given by a new non-parametric method.
Keywords:
optimal histogram class interval, non-stationary distributions, time series
Publication language:russian, pages:26
Research direction:
Mathematical modelling in actual problems of science and technics