The Use of Time Series to Improve the Prediction Reliability of the Statistical Characteristics of a Controlled Parameter

Abstract

Based on the analysis of time series compiled from the values of a controlled parameter, we propose options for constructing vectors (vector-neighbors), close to the starting vector. It is shown that the use of the coordinates of the constructed vector-neighbors in conjunction with the coordinates of the starting vector allows one to increase the sample sizes for determining the statistical characteristics during prediction of parameter drift and to increase the reliability of the estimates of these characteristics.