Abstract
Analytical expressions were derived for the expectation of the autocorrelation function (ACF) corresponding to lowfrequency least squares fits and deviations from them in the case of a limited number of observationsN. A vector of values of the autocovariation function may be obtained by multiplication of aN . NmatrixZ(dependent on concrete basic functions used for trend determination) by a vector of the unbiased ACF. Because much computational time is needed to obtain such a matrix, its components are to be computed once for concreteNand basic functions, and then stored as a file. An algorithm is proposed for determining the contribution of the correlated signal to the ‘signal+noise’. The expressions are written for general form of the ACF, and illustrated by the application to autoregressive models. The statistical properties of the model parameters are studied. The method is applied to cataclysmic binaries AM Her and TT Ari. The precise expressions allow us to obtain reliable results and to avoid misinterpretation being possible when using the approximate methods.

This publication has 13 references indexed in Scilit: