Comparative Analysis of the Various Methods Stability in Evaluation of the Bilinear Autoregression Model Parameters

Abstract

The purpose of this work is to compare various methods in evaluating parameters of a bilinear autoregressive model. Least squares estimate, least absolute deviations estimate and estimate based on the Huber function were used as the parameter estimates. Computer simulation was introduced to study the indicated estimates precision depending on probability distribution of the bilinear autoregressive model upgrading process. Probable distribution of the upgrading process was simulated by normal and uniform distributions, Student distribution with various degrees of freedom, Laplace distribution (double exponential distribution), and Tukey distribution known as the polluted normal distribution. Their mean square error served as the evaluation precision measure. Results of the conducted computational experiment showed that precision of three methods used in evaluating parameters of the bilinear autoregressive series significantly depended on probability distribution of the model upgrading process. In particular, it concerns the number of degrees of freedom of the Student distribution, as well as the Tukey distribution pollution share and amount. If the upgrading process possesses normal and uniform distributions, Student’s distribution with sufficiently high number of degrees of freedom, the least squares method works more efficiently. Estimate based on the Huber function and the least absolute deviations estimate are becoming more efficient compared to the least squares estimation for the Laplace distribution, with a decrease in the number of degrees of freedom --- for the Student distribution, and with an increase in the pollution share and amount --- for the Tukey distribution

Please cite this article in English as:

Andreychik N.L., Goryainov V.B. Comparative analysis of the various methods stability in evaluation of the bilinear autoregression model parameters. Herald of the Bauman Moscow State Technical University, Series Natural Sciences, 2022, no. 6 (105), pp. 4--16 (in Russ.). DOI: https://doi.org/10.18698/1812-3368-2022-6-4-16

References

[1] Stefansson P., Fortuna J., Rahmati H., et al. Chapter 2.12 --- Hyperspectral time series analysis: hyperspectral image data streams interpreted by modeling known and unknown variations. Data Handl. Sci. Technol., 2020, vol. 32, pp. 305--331. DOI: https://doi.org/10.1016/B978-0-444-63977-6.00014-6

[2] Ismail M.I., Ali H., Syed Yahaya S.S. The evaluation of robust outlier detection procedure in bilinear (1,0,1,1) model. Test, 2019, vol. 81, no. 11-12, pp. 4001--4007.

[3] Zhang Y., Li R., Min Q., et al. The controlling factors of atmospheric formaldehyde (HCHO) in Amazon as seen from satellite. Earth Space Sci., 2019, vol. 6, iss. 6, pp. 959--971. DOI: https://doi.org/10.1029/2019EA000627

[4] Liu P.F., Peng X.Q., Guo Z.Y., et al. Viscoelastic bilinear cohesive model and parameter identification for failure analysis of adhesive composite joints. Compos. Struct., 2019, vol. 224, art. 111016. DOI: https://doi.org/10.1016/j.compstruct.2019.111016

[5] Whitney R. On the use of a bilinear noise model to recover residual displacement from strong motion recordings. Geotechnical Earthquake Engineering and Soil Dynamics V: Seismic Hazard Analysis, Earthquake Ground Motions, and Regional-Scale Assessment, 2018, pp. 559--568. DOI: https://doi.org/10.1061/9780784481462.054

[6] Shimi M., Najjarchi M., Khalili K., et al. Investigation of the accuracy of linear and nonlinear time series models in modeling and forecasting of pan evaporation in IRAN. Arab. J. Geosci., 2020, vol. 13, no. 2, art. 59. DOI: https://doi.org/10.1007/s12517-019-5031-7

[7] Raymer J., Wisniowski A. Applying and testing a forecasting model for age and sex patterns of immigration and emigration. Popul. Stud., 2018, vol. 72, iss. 3, pp. 339--355. DOI: https://doi.org/10.1080/00324728.2018.1469784

[8] Goryainov V.B., Goryainova E.R. Robust estimation in threshold autoregression. Herald of the Bauman Moscow State Technical University, Series Natural Sciences, 2017, no. 6 (75), pp. 19--30 (in Russ.). DOI: http://dx.doi.org/10.18698/1812-3368-2017-6-19-30

[9] Goryainov A.V., Goryainov V.B. M-Estimates of autoregression with random coefficients. Autom. Remote Control, 2018, vol. 79, no. 8, pp. 1409--1421. DOI: https://doi.org/10.1134/S0005117918080040

[10] Goryainov V.B., Goryainova E.R. Asymptotic properties of the sign estimate of autoregression field coefficients. Autom. Remote Control, 2015, vol. 76, no. 3, pp. 419--432. DOI: https://doi.org/10.1134/S0005117915030066

[11] Liu J. On stationarity and asymptotic inference of bilinear time series models. Stat. Sin., 1992, vol. 2, no. 2, pp. 479--494.

[12] Cavanaugh J.E., Neath A.A. The Akaike information criterion: background, derivation, properties, application, interpretation, and refinements. WIREs Computational Statistics, 2019, vol. 11, iss. 3, art. e1460. DOI: https://doi.org/10.1002/wics.1460

[13] Attetkov A.V., Zarubin V.S., Kanatnikov A.N. Vvedenie v metody optimizatsii [Introduction to optimization methods]. Moscow, Finansy i statistika Publ., 2008.

[14] Beck A., Sabach S. Weiszfeld’s method: old and new results. J. Optim. Theory Appl., 2015, vol. 164, no. 1, pp. 1--40. DOI: https://doi.org/10.1007/s10957-014-0586-7

[15] Huber P.J., Ronchetti E.M. Robust statistics. Wiley, 2009.

[16] Maronna R.A., Martin R.D., Yohai V.J., et al. Robust statistics. Theory and methods (with R). Wiley, 2019.