Comparison between classical and robust estimation methods for regression model parameters in the case of incomplete data
Keywords:
R-Estimators; L estimators; EM; W estimators; incomplete data; robust methodsAbstract
The problem of complete data is considered one of the most important problems that hinder statistical analysis. Therefore, this problem must be solved by finding sound solutions to it by using some methods that lead to accurate results or close to accuracy. The research aims to compare methods of statistical data analysis. Estimating incomplete data using the filling algorithms R-Estimators, L estimators, EM, and and the W method and the case of the normal distribution of data according to patterns, machines, and the method of losing these observations. A comparison was made between the robust methods and the maximum Likelihood method, and the efficiency of the above-mentioned methods was noted. The imitation method was used in addition to practice. the data on the function variables, which represent the demand for cash and its relationship with the gross domestic product, government consumer spending, and the consumer price index, were discussed for the period from 2000-2022.The process was to compare these methods, and the results were drawn to reach the experimental side , through the experimental and applied part, it is preferable to use robust methods over classical methods in the case of incomplete data.
References
Basu, S. & Heitjan, D. F. (2006) ”Distinguishing ’Missing at Random’ & ’Missing completely at Random’ The JASA, vol. 23, No. 4, p. 123- 132.
Beal, M.J., & Ghahramani, Z. (2016) ”The variational Bayesian EM Algorithm for Incomplete Data: with Application to Scoring Graphical Model structures”, www.stat.cmu.edu/fienberg/stat36-756-06.html.
Campbell, N. A (1999) ”Robust procedures in Multivariate Analysis I: Robust covariance Estimation” Applied statistics, vol. 23, P. 213- 224.
David, T. O (2004) ”Applied Robust statistics” Chapman & Hall, New York.
Dempster, A. P. & Larid, N. M. & Rubin, D. B. (1977) ”Maximum Likelihood from Incomplete data via the EM Algorithm” J. Royal Statist. Soc. Ser. B, vol. 11, P. 1-22.
Honaker, J.; Joseph, A.; King, G. & Scheve, K. (2001) ”Analyzing Incomplete political science data: An Alternative Algroithm for Multiple Imputation”American political science review, vol. 68, No. 2, P. 123-137.
Jank, W. (2006) ”The EM Algorithm, it’s stochastic Implementation and Global Optimization: some challenges and opportunitiesfor OR”, www.rhsmith.umd.edu/faculty/wjank/.../page1118.htm.
Jurckova, J. & Sen, P. K. (1996) ”Robust statistical procedures; Asymptotic and Interrelations” John Wile & Sons, New York.
Lin, C. & Rubin, D. B. (2002) ”Maximum Likelihood Estimation of Factor Analysis using the ECME Algorithm with complete and Incomplete Data”, statistica sinica 8, p. 437-446.
Little, R. J. A. & Rubin, D. B. (2002) ”Statistical Analysis with Missing Data”, New York: John. Wiley.
Little, R. J. A. & Rubin, D. B (2003)” statistical analysis with missing data”, 2nd ed., John Wile & sons, New York.
Maronna, R. A. (1976) ”Robust M- Estimators of Multivariate location and scatter” The Annals of statistics, vol. 3, P. 33- 48.
Morrison, D. F. (1971) ”Expectation and variances of Maximum likelihood Estimates of the Multivariate Normal Distribution parameters with missing data”, JASA, vol. 53, P. 326- 339.
Peter, J. H (1981) ”Robust statistic” John Wile & Sons, New York.
Roche. A. (2006) ”EM Algorithm and variants: an informal tutorial”, ftp://ftp.cea.fr/pub/dsv/madic/pubils/rocheem.
Rubin, D. B. (1974) ”Characterizing the Estimation of parameters in Incomplete- Data problems” Journal of the American statistical association.33 Sehie Park et. al., Adv. Theory Nonlinear Anal. Appl. 7 (2023), 17–34.
Bick el,D.R.,(2002). ”Robust and efficient estimation of the mode of continuousdata ”http://www.mcg.edu/research/biostat/b
Daniel P. Normolle .(2003).”An Algorithm for Robust nonlinearAnalysis of Radioimmunoassay and Othre Bioas says ”Ann Arbor,MI 48109-2029 , 313-936-1013.
