Comparison of time series using subsampling

Existing procedures for the comparison of stationary time series are based on tests which require either model estimation or spectral estimation. In most cases these procedures are applicable to pairs of time series that are assumed to be generated independently of each other. A procedure that is based on subsampling techniques, and is free from either model or spectral estimation is proposed. It is applicable to pairs of time series that may or may not be assumed to be independently generated. This procedure for which consistency is established, involves the use of a test based on the Euclidean distance between the autocorrelation functions of two time series. The performance of the test is evaluated using a Monte Carlo study. This study reveals that the test performs reasonably well and is competitive with existing procedures for the comparison of stationary time series. The test is applied to a set of observed financial time series.