ANALYTICAL BAYES ESTIMATOR AND DISTRIBUTION FOR OUTLIER INFESTED TIME SERIES DATA

摘要

The problem of specifying time domain model in the presence of outlier is considered using information from all realizations, and an analytical Bayes estimator is proposed for outlier infested time series data. This approach is practicable and the Bayes estimator is considered to have more precision over the conventional classical outlier series estimator. The conditional distribution of the number of outlier-infested series given an outlier-free series is proposed to accommodate the problem of overlapping distributions for a given sequence of series.