Robustification of Sequential Statistical Decision Rules for Stochastic Data Flows Analysis

Abstract

In data analysis the issues of statistical decision making on parameters of observed stochastic data flows are important. To solve the relevant problems, here sequential statistical decision rules are used. The sequential statistical decision rules traditionally used loose their performance optimality in situations that are common in practice, when the hypothetical model is distorted. Here the robustified sequential decision rules are constructed for three models of observation flows: independent homogeneous observations; observations forming a time series with a trend; dependent observations forming a homogeneous Markov chain.