Detecting Over-Influential Observations in Time Series
Bernard Lefrançois · Biometrika · 1991
The purpose of this paper is to develop a tool for identifying over-influential observations in time series when they are viewed in the time domain. We present a method for obtaining various measures of influence for the autocorrelation function, as well as thresholds for declaring an observation over-influential. An example of the use of these thresholds is also presented.