Bootstrap assisted goodness of fit tests in the frequency domain

摘要: The validity of a stationary time series model may be measured by the goodness fit spectral distribution function. Anderson (Technical Report 27, 1991; Technical 309, 1995; Stanford University) has worked out closed-form characteristic functions for Cramer–von Mises criterion general linear processes, under condition that all values parameters are specified. asymptotic approach is not easily implemented and usually requires case analysis. In this paper we propose bootstrap goodness-of-fit test in frequency domain. By properly resampling residuals, can consistently estimate p many weakly dependent semiparametric models with unspecified parameter values. This content main theorem try to explain. A group simulations conducted, showing consistent significance level good power. special tests applied lynx data reveal structure unexplained AR(1) fitted Tong (J. R. Stat. Soc. 140 (1977), 432–36). possible generalization application financial analysis also discussed.