STUDY OF SOME KINDS OF ALMOST UNBIASED ESTIMATOR AND PRINCIPLE COMPONENT ESTIMATOR FOR REGRESSION MODEL

Abstract

In this research, two types of bias estimator were studied in linear regression model which are; (almost unbiased generalized ridge estimator, almost unbiased two-parameter estimator and Modified ridge-type estimator) and (The (r-k) class estimator and modified (r-k) class ridge regression estimator) as a method of repressing the multicollinearity problem on parameter estimation in multiple linear regression models. Also, a simulation analysis was used to test the relative efficiency of certain types of biased estimators as well as the thirty-nine proposed estimated ridge parameters (k) that have been shown in the literature. Moreover, the mean square error was also assigned to study the quality of those estimators in different circumstances and for different correlations. Finally, a practical example was applied to illustrate the obtained results. All proposed estimators of (k) are, according to the results, superior to ordinary least squared estimator (LSE), but there is no ‘optimal’ estimator guarantee that can come out, and the best estimator option will depend on the study conditions.