Performances of the hybrid-sign correlation coefficient estimator for Gaussian stationary processes

The hybrid-sign estimator is a method of estimating the autocorrelation coefficient for Gaussian stationary processes based on sums. In this correspondence the features of such an estimator are examined, both for uncorrelated pairs of samples, and for correlated pairs with an assigned autocorrelation function of the process. A generalized estimator is considered, by varying the spacing between the pairs of samples. The performances, in terms of bias, covariance, and variance, are given in simple approximate formulas, and compared in comprehensive diagrams to the direct estimator, based on sums of products, and the polarity coincidence estimator, based on counters.