Bootstrap for Maximum Likelihood Estimates of PARMA Coefficients
In this chapter we use bootstrap techniques to estimate empirical distributions of parameter estimates for PAR sequences determined by maximum likelihood techniques. The parameters are not the periodic autoregression parameters, but are the coefficients in the Fourier series representing the parameters. We compare two different bootstrap techniques, IID and GSBB, applied to the residuals of the maximum likelihood estimation. The IID method seems a little better, which is not a surprise since the conditions for the GSBB are not completely satisfied. We expect these method to also work satisfactorily for full PARMA estimations, where both PMA and PAR terms are present in the model.
Research of Anna Dudek was partially supported by the Polish Ministry of Science and Higher Education and AGH local grant.
- Dudek AE, Hurd H, Wójtowicz W (2013) PerARMA: pacakge for periodic time series analysis, R package version 1.5. http://cran.r-project.org/web/packages/perARMA/
- Dudek AE, Leśkow J, Politis DN, Paparoditis E (2014) A generalized block bootstrap for seasonal time series, article first published online: 27 NOV 2013, doi: 10.1002/jtsa.12053
- Hurd HL, Miamee AG (2007) Periodically correlated random sequences: spectral. Theory and practice. Wiley, New YorkGoogle Scholar
- Politis DN (2001) Resampling time series with seasonal components, in Frontiers in data mining and bioinformatics. In: Proceedings of the 33rd symposium on the interface of computing science and statistics. Orange County, California, 13–17 June, pp 619–621Google Scholar