A hypothesis test using bias-adjusted AR estimators for classifying time series in small samples

摘要

Purpose: To propose a test of hypothesis for classifying stationary time series based on the bias-adjusted estimators of the fitted autoregressive model. Summary: Classification of time series analysis is a interest in many fields such as economics, medical diagnosis data like ECC data, EMC data, data from biological studies, seismological studies, finance environmetrics etc. There are three classification methods that are used for this purpose namely, raw-data-based, feature based, and model-based. The first two work with raw data or features associated with the data such as autocorrelations, wavelets etc and the model- based method uses combinations of various underlying distributions that generate the data. Auto-regressive moving average (ARMA) models are mostly used for comparing different time series fitted to the data. To achieve desirable accuracy, bias-adjusted AR estimators are to be considered when the sample size is small. This paper proposes a hypothesis test based on bias-adjusted AR estimator that can be used when the time series are not independent. (51 refs.)