Estimating polynomial models with errors in variables without additional information

V. I. Denisov ; A. Yu. Timofeeva ; E. A. Khailenko

Sibirskij žurnal industrialʹnoj matematiki, Tome 19 (2016) no. 3, pp. 15-27.

Résumé

We consider the problem of estimating a polynomial model with classical error in the input factor in the functional case. The nonparametric method of estimation of structural dependencies does not use additional information but is extremely hard computationally and requires a large sample size. That is why we propose a number of easier approaches. The first approach is based on the preliminary estimation of the Berkson error variance under the assumption of its normality for a piecewise-linear model. The so-obtained estimate is used to calculating the parameters of the polynomial by the methods of general and adjusted least squares. In the case when the error distribution deviates from the normal distribution, we develop a second method, the adaptive estimation method, based on the universal lambda-distribution. The proposed approaches were developed for solving the problem of analyzing the level of knowledge.

Détail

Keywords: model with errors in both variables, method of generalized least squares, method of adjusted least squares, maximum likelihood method, adaptive method, generalized distribution.

Bibliographie

[1] Fuller W. A., Measurement Error Models, John Wiley and Sons, NY, 1987 | MR | Zbl

[2] Carroll R. J., Ruppert D., Stefanski L. A., Measurement Error in Nonlinear Models, Chapman and Hall, London, 1995 | MR | Zbl

[3] Timofeeva A. Yu., Avrunev O. E., “Lokalno vzveshennoe vosstanovlenie strukturnykh zavisimostei v zadache analiza uspevaemosti”, Dokl. AN VSh RF, 22:1 (2014), 135–146 | MR

[4] Poldin O. V., “Prognozirovanie uspevaemosti v vuze po rezultatam EGE”, Prikl. ekonometrika, 2011, no. 1, 56–69

[5] Khavenson T. E., Soloveva A. A., “Svyaz rezultatov Edinogo gosudarstvennogo ekzamena i uspevaemosti v vuze”, Voprosy obrazovaniya, 2014, no. 1, 176–199

[6] Kobrin J. L., Patterson B. F, Shaw E. J., Mattern K. D., Barbuti S. M., Validity of the SAT for predicting first-year college grade point average, College Board Research Report No 2008-5, The College Board, NY, 2008

[7] Timofeeva A., “Orthogonal regression for nonparametric estimation of errors-in-variables models”, J. Math., Computat. Phys., Quantum Engrg., 8:8 (2014), 470–474 | MR

[8] Marini J. P., Mattern K. D., Shaw E. J., Examining the linearity of the PSAT/NMSQTRFYGPA relationship, College Board Research Report No 2011-7, The College Board, NY, 2011

[9] Rousseeuw P. J., Van Driessen K., Computing LTS Regression for Large Data Sets, Univ. of Antwerpen, Antwerpen, 1999

[10] Timofeev V. S., Khailenko E. A., “Adaptivnoe otsenivanie parametrov regressionnykh modelei s ispolzovaniem obobschennogo lyambda-raspredeleniya”, Dokl. AN VSh RF, 2010, no. 2(15), 25–36

[11] Kendall M., Styuart A., Statisticheskie vyvody i svyazi, Nauka, M., 1973

[12] Van Huffel S., Vandewalle J., The Total Least Squares Problem: Computational Aspects and Analysis, SIAM, 1991 | MR | Zbl

[13] Greshilov A. A., Stakun V. A., Stakun A. A., Matematicheskie metody postroeniya prognozov, Radio i svyaz, M., 1997

[14] Denisov V. I., Timofeeva A. Yu., Khailenko E. A., Buzmakova O. I., “Ustoichivoe otsenivanie nelineinykh strukturnykh zavisimostei”, Sib. zhurn. industr. matematiki, 16:4 (2013), 47–60 | MR | Zbl

[15] Cheng C.-L., Schneeweiss H., “Polynomial regression with errors in the variables”, J. Royal Statist. Soc. Ser. B, 60 (1998), 189–199 | DOI | MR | Zbl

[16] Schennach S. M., “Estimation of nonlinear models with measurement error”, Econometrica, 72:1 (2004), 33–75 | DOI | MR | Zbl

[17] Hausman J., Newey W. K., Powell J. L., “Nonlinear errors in variables: Estimation of some Engel curves”, J. Econometrics, 65:1 (1995), 205–233 | DOI | MR | Zbl

[18] Timofeeva A. Yu., “K probleme endogennosti potrebitelskikh raskhodov pri otsenivanii sistemy sprosa domashnikh khozyaistv”, Prikl. ekonometrika, 37:1 (2015), 87–106

[19] Schennach S. M., Hu Y., “Nonparametric identification and semiparametric estimation of classical measurement errormodels without side information”, J. American Statistical Association, 108:501 (2013), 177–186 | DOI | MR | Zbl

[20] Timofeeva A. Yu., Buzmakova O. I., “Poluparametricheskoe otsenivanie zavisimostei mezhdu stokhasticheskimi peremennymi”, Nauchn. vestn. NGTU, 49:4 (2012), 29–37

[21] Wang L., “Estimation of nonlinear models with Berkson measurement errors”, Annals of Statistics, 32:6 (2004), 2559–2579 | DOI | MR | Zbl

[22] Huwang L., Huang Y. H. S., “On errors-in-variables in polynomial regression – Berkson Case”, Statistica Sinica, 10 (2000), 923–936 | MR | Zbl

[23] Karian Z. A., Dudewicz E. J., Fitting Statistical Distributions: the Generalized Lambda Distribution and Generalized Bootstrap Methods, CRC Press LLC, NY, 2000 | MR | Zbl

[24] Lakhany A., Mausser H., “Estimation the parameters of the generalized lambda distribution”, ALGO Research Quarterly, 3:3 (2000), 27–58

[25] Stasyshin M. V., Avrunev O. E., Afonina E. V., Lyakh K. N., “Informatsionnaya sistema universiteta: opyt sozdaniya i tekuschee sostoyanie”, Otkrytoe i distantsionnoe obrazovanie, 46:2 (2012), 9–15

[26] Rousseeuw P. J., Croux C., “Alternatives to the median absolute deviation”, J. Amer. Statist. Assoc., 88:424 (1993), 1273–1283 | DOI | MR | Zbl