estimating autocorrelation function using singular spectrum analysis

Straightforward manipulation shows that the sum of the sample autocorrelationfunction over the lags 1 to t − 1 equals to -0.5 for all time series of lengtht. there are some alternatives in the literature, however those are not common inpractice. in this paper, we provide a new approach for estimating the autocorrelationfunction. this approach uses singular spectrum analysis which is a non-parametrictechnique for time series analysis. the paper utilizes a simulation study to illustratethe performance of the new approach. the results suggest that further improvementto the sample autocorrelation is possible and the new method provides an attractivealternative to the classical approach.