Abstract:
The objective of this research is to compare multiple regression coefficients estimating methods with existence of multicollinearity among independent variables. The estimation methods are Empirical Bayes Ridge regression method (EB), Hierarchical Empirical Bayes Ridge regression method (HB) and Decomposed Empirical Bayes Ridge regression method (DB). The Average Mean Square Error (AMSE) was used as the criteria in this research. The criterion of comparison is the Ratio of Different Average Mean Square Error (RDAMSE). This research used 7 and 9 independent variables. The number of remaining eigenvalues are 3 and 5. The size of the samples was vary, which composed of 30, 50, 70 and 100 samples. The distribution of error is normal distribution with mean equal to 0 and standard deviation equal to 1, 5 and 10, respectively. The level of correlations among the independent variables could be classified into 3 levels for which low levels equal to (0.10, 0.30), middle levels equal to (0.50, 0.70) and high level equal to (0.80, 0.90). The study used the Monte Carlo Simulation method. The experiment was repeated 1,000 times under each situation. The analyzed results of the data was demonstrated as follow. CASE 1 The number of independent variables is 7. In all cases, DB method has the smallest AMSE and HB method has a smaller AMSE than the EB method, respctively. The AMSE decreases when sample size increases but it increases when level of correlation or standard deviation or the number of remaining eigenvalues increases. The RDAMSE decreases when sample size increases but it increases when level of correlation or standard deviation or the number of remaining eigenvalues increases. CASE 2 The number of independent variables is 9. The result is the same as the case 1. The AMSE varies with, most to least, respectively, level of correlations, standard deviation, number of independent variables and number of remaining eigenvalues but converses to sample sizes.