Multiple Regression Using R: A Comparison of All Possible Subset Regression, Beta Weights, Structure Coefficients, and Commonality Analysis

• In Event: 46.063 - Multiple Regression: Principles and Practices

Abstract

One main concern when conducting multiple regression analysis is the selection of predictor variables. A common method of selecting predictor variables is testing the significance of the beta weights. One may use the unstandardized or standardized beta weights and should be able to conclude which are the significant predictor variables. Another method would be to test the significance of the structure coefficients. There is also a method called commonality analysis that gives all possible subsets of variable combination models to help identify multicollinearity and suppressor variable effects. Using the R software package, yhat, all of these variable selection methods can be compared. Each method will be discussed, computed, and compared in reference to multiple regression modeling.

Author

• Brigette Winkles, The University of Alabama