Parametric polyspectrum density estimation using the bootstrap method

A method for obtaining the statistical distribution of parametric polyspectra when applied to a single set of short data record is presented. The method applies model-based bootstrap to a time-series data to obtain the statistical distribution of the coefficients of the approximating process of autoregressive moving average (ARMA) type with known order (p,q)(p,q). As the polyspectral estimate depend on these coefficients, relevant statistical information of the estimated polyspectrum is obtained from the distribution function of the coefficients. Examples of a simulated and real data are provided to illustrate the method.