Prewindowed and postwindowed methods for bispectrum estimation via AR modelling

Two new autoregressive parametric methods for the bispectrum estimation of a stationary process are presented. These are an extension of the prewindowed (PREW) and postwindowed (POSW) methods for the estimation of a power spectrum via AR modelling. A variant of each of these methods, based on the segmentation of process data into coupled records, is also presented. Simulations show that the POSW and PREW methods produce similar results to those obtained using the third order recursion (TOR) and constrained third-order mean (CTOM) methods, respectively, and that their modifications provide a more precise estimation than that obtained using the TOR and CTOM methods of the AR-model parameters.