Selecting regression model

• Autorzy: Visek, J. Á.
• Języki publikacji: EN

Abstrakty

EN

A new tool for the identification of regression model is proposed and its properties are established. The key importance of the new tool is that it is able to solve still not very well-known problem of diversity of estimates, as described in Višek [22] and [25]. Main idea of the proposal is as follows. Having evaluated an estimate of regression coefficients for given data, the data are partitioned into two disjoint subsets (e.g. by a geometric rule applied in the factor space). Then for each subset of corresponding residuals we evaluate the estimate of their density, e.g. the kernel one. If the estimate of regression model is “near to the true model”, the density of disturbances is the same in the both subsets, and hence also the estimates of density of residuals are approximately equal each to other. Therefore, finally, the estimates of density are compared by means of the weighted Hellinger distance. It implies that the significant difference between the estimates of density indicates that the given estimate of the regression model is not near to the “true” model or, in other words, that it is not “adequate” for the data. In the case when we have at our disposal more estimates of the regression model, and especially when the estimates are considerably different (each from other), the test statistic may be also used for selecting the estimate of the regression model. We just accept the estimate with the smallest weighted Hellinger distance. The result of the paper is illustrated by two simple numerical examples demonstrating especially the sensitivity of the test statistic to the difference between the estimates of density.

Słowa kluczowe

EN

weighted Hellinger distance; diagnostic of model; choice of model; diversity of robust estimates

• Czasopismo: Probability and Mathematical Statistics
• Rocznik: 2001
• Tom: Vol. 21, Fasc. 2
• Strony: 467--492
• Opis fizyczny: Bibliogr.27 poz., tab., wykr.

Twórcy

autor

Visek, J. Á.

• Department of Macroeconomics and Econometrics, Institute of Economic Studies, Faculty of Social Science and Automation Charles University ,
• Department of Stochastic Informatics Institute of Information Theory and Automation, Academy of Sciences of Czech Republic, Opletalova ulice 26, CZ-11000 Prague 1, Czech Republic

Bibliografia

• [1] M. Abramowitz and L A. Stegun, of Mathematical Functions with Formulas, Graphs and Mathematical Tables, Dover Publications, 1964.
• [2] P. Boček and P. Lachout, Linear programming approach to LMS-estimatiom in: Memorial volume of “Compta. Statist. Data Anal” devoted to T. Havránek 19 (1995), pp. 129-134.
• [3] P. Boček and J. Á. Visek, On one-step M-estimates, in: Transactions of the Fifth Prague Symposium on Asymptotic Statistics, M. Hušková and P. Mandl (Eds.), Physica Verlag, 1994f pp. 395 202.
• [4] L. Breiman, Probabiliy, Addison-Wesley, London 1968.
• [5] M. Csörgő and P. Révész, Strong Approximations in Probability and Statistics, Akadémiai Kiadi, Budapest 1981.
• [6] t. R. Hampel P. J. Rousseeuw, E, M. Ronchetti and W, A. Stahel Robust Statistics - The Approach Based on influence Functions, Wiley, New York 1986.
• [7] R. V, Hogg, Adaptive robust procedures: A partial review and some suggestions for future applications and theory (with discussion), J. Amer. Statist. Assoc, 69 (1974), pp, 909-927.
• [8] R. M. Humphreys, Studies of luminous stars in nearby galaxies. L Supergiants and 0 stars in the milky way, Astrophys. J. Suppt Ser. 38 (1978), pp. 309-350.
• [9] J. Jurečková, Consistency of M-estimators in linear model generated by nonmonotone and discontinuous ψ-fuctions, Probat, Math. Statist. 1.0 (1988), pp. 1-10.
• [10] X Jurečková, Consistency of M-estimators of vector parameters, in; Proceedings of the Fourth Prague Symposium on Asymptotic Statistics, Charles University, Prague 1988, pp. 305-312.
• [U] X Jurečková and S, Portnoy, Asymptotics for one-step M-estimators in regression with application to combining efficiency and high breakdown point. Comm, Statist. A 16 (1988), pp. 2187-2199.
• [12] M. Markatou, W. A. Stahel and E. M. Ronchetti, Robus M-type testing procedures for linear models, in; Directions in Robust Statistics and Diagnostics, W. Stahel and S. Weisberg (Eds.), Springer, New York 1991, pp. 201-220.
• [13] R. A. Maronna and V. J. Yohai, Asymptoic behaviour of general M-estimates for regression and scale with random carriers, Z. Wahrsch. verw. Gebiete 58 (1981), pp. 7 20.
• [14] R. D. Martin, V. J Yohai and R. H. Zamar, Min-max bias robust regression, Ann. Statist 17 (1989), pp. 1608-1630.
• [15] P, J. Rousseeuw and A, M. Leroy, Robust Regression and Outlier Detection, Wiley, New York 1987.
• [16] S. S. Shapiro and M. B. Wilk, A comparative study of various tests for normality, J. Amer. Statist. Assoc. 63 (1968), pp. 1343 1372.
• [17] J. Á, Víšek, Stability of regression model estimates with respect to subsamples. Comprit. Statist, 7 (1992), pp.183-203.
• [18] J Á. Víšek, Adaptive maximum-likelihood-like estimation in linear model. Part X Consistency, Kybernetika 28 (1992), pp. 357-382. Part II. Asymptotic normality, Kybernetika 28 (1992), pp. 454-471.
• [19] J, Á. Víšek, On the role of contamination level and the least favourable behaviour of gross-error sensitivity, Probab. Math. Statist. 14 (1993), pp. 173-187.
• [20] J. Á. Víšek, A cautioanry note on the method of Least Median of Squares reconsidered, in: Transactions of the Twelfth Prague Conference on Information Theory, Statistical Decision Functions and Random Processes, Prague 1994, pp. 254-259.
• [21] J. Á. Víšek, High robustness and an illusion of truth, in.; Transactions of ROBUST’94, published bv JČMF (Union of Czech Mathematicians), J. Antoch and. G. Dohnal (Eds], 1994, pp. 172-185.
• [22] J. Á, Víšek, On high breakdown point estimation, Comput. Statist, 11 (1996), pp. 137 .146.
• [23] J. Á. Víšek, Contamínation level and sensitivity of robust tests, in: Hnndbook of Statistics, VoL 15, G. S. Maddala and C. R. Rao (Eds.), Elsevier Science B. V., Amsterdam 1997, pp. 633-642.
• [24] J A. Víšek, The least trimmed squares - random carriers. Bulletin of the Czech Econometric Society 10 (1999), pp. .1. 30.
• [25] J A. Víšek, On the diversity of estimates, Comput. Statist. Data Anal 34(2000) pp. 67-89.
• [26] J A. Víšek, Sensivity analysis of M-estimates of nonlinear model: influence of data subsets, Ann. Inst. Statist, Math., to appear.
• [27] V. J. Yohai and R. H. Zamar, A minimax-bias property of the least α-quantile estimates, Ann. Statist. 21 (1.993), pp. 1824-1842.