zbMATH — the first resource for mathematics

Computer algebra derivation of the bias of linear estimators of autoregressive models. (English) Zbl 1115.62093
An AR(2) model is considered, \(z_t=\varpi_1z_{t-1}+\varphi_2 z_{t-2}+\varepsilon_t\), with i.i.d. Gaussian zero mean innovations \(\varepsilon_t\). For the Burg estimates \(\hat\varphi_i\) by the observations \(z_j\), \(j=1,\dots,n\), the principal part of the bias is derived, e.g., \[ \lim_{n\to\infty} n E(\hat\varphi_2-\varphi_2)=-(1+3\varphi_2)/n. \] Analogous results are obtained for estimates by observations of \(z_t+m\), where \(m\) is a constant to be estimated. For AR(3) the authors describe the computer derivation (via symbolic computations) of analogous formulas. E.g., for \(\varphi_3\) such formula includes 1,745,350 indivisible subexpressions in terms of roots of the characteristic equation. Numerical results for comparison of bias for the Burg, least squares and Yule-Walker estimates are presented. The authors’ conclusion is that Burg and LS estimates have bias much better than the Yule-Walker ones when the parameters are near the nonstationary boundary. Simulation results are presented.

62M10 Time series, auto-correlation, regression, etc. in statistics (GARCH)
65C60 Computational problems in statistics (MSC2010)
62M09 Non-Markovian processes: estimation
68W30 Symbolic computation and algebraic computation
Full Text: DOI arXiv
[1] Andrews D. F., Journal of the Royal Statistical Society B 55 pp 613– (1993)
[2] Andrews D. F., Symbolic Computation for Statistical Inference (2000) · Zbl 0995.62002
[3] Bartlett M. S., Journal of the Royal Statistical Society Supplement 8 pp 27– (1946)
[4] Cook P., The American Statistician 49 pp 70– (1995)
[5] Currie I. D., Applied Statistics 44 pp 379– (1995)
[6] Fuller W. A., Introduction to Statistical Time Series (1996) · Zbl 0851.62057
[7] Kendall M. G., Biometrika 41 pp 403– (1954)
[8] Lysne D., Biometrika 74 pp 200– (1987)
[9] Marriott E. H. C., Biometrika 41 pp 390– (1954)
[10] Percival D. B., Spectral Analysis For Physical Applications (1993) · Zbl 0796.62077
[11] Rose C., The Statistician 49 pp 229– (2000)
[12] Rose C., Mathematical Statistics with Mathematica (2002) · Zbl 0989.62001
[13] DOI: 10.1023/A:1026514117160
[14] TjOstheim D., Biometrika 70 pp 389– (1983)
[15] White J. S., Biometrika 48 pp 85– (1961)
[16] Wolfram S., The Mathematica Book (2003)
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. It attempts to reflect the references listed in the original paper as accurately as possible without claiming the completeness or perfect precision of the matching.