Assessing the performance of change-point detection based on the SIC procedure for non Gaussian and correlated data - a simulation study

Abstract

A simulation study is conducted in order to assess the impact of autocorrelation and non-normality on change-point detection when it is performed by the Schwarz Information Criterion (SIC) approach. Change-point detection in environmental data is usually fulfilled in the presence of time correlation or non-Gaussian distribution. This work analyses a set of scenarios with the purpose of evaluating the SIC performance when the assumptions are not guaranteed.