Advanced Regression Analysis: Evaluating Collinearity, Model Selection,and Predictive AccuracyDiscussion Questions: 15.26How can you evaluate whether collinearity exists in a model?The collinearity in a model is a condition in multiple linear regression when few of theindependent variables are highly correlated with each other causing the coefficients to beunstable.If collinearity exists, collinear variables do not provide unique information, and it becomesdifficult to separate the effects of such variables on the dependent variable.One method of measuring collinearity is to determine the variance inflationary factor (VIF) foreach independent variable. Higher VIFs indicate the presence of collinearity in the model.Discussion Questions: 15.27What is the difference between stepwise regression and best-subsets regression?Both of these regression model select a regression model from the subset of variables. To beclearer, suppose we have large number of independent variables and we need a good model withas less number of variables as possible. In that situation, both of these technique would workreally well to select subset model.However the basic difference between them is the steps used. In stepwise regression theindependent variable having highest effect is selected 1stthen the variable with a little lesssignificant effect and so on. By adding and removing the variables it finds out the best subsetmodel. However as it only considers the best or most significant independent variable so it doesnot construct all the possible models.And in best-subsets regression all the possible model is constructed for each number ofindependent variables. Then we need to select the best model based on the output. Clearly thenumber of model is much larger in this case than thestepwise regression model.Preview Mode
This document has 5 pages. Sign in to access the full document!