Chapter 13 Multiple Regression Analysis.PPT

Chapter 13 Multiple Regression Analysis General Objectives: In this chapter, we extend the concepts of linear regression and correlation to a situation where the mean value of a random variable is related to several independent variables—x1, x2, x3, …, xk—in models that are more flexible than the straight-line model of Chapter 12. With multiple regression analysis, we can use the information provided by the independent variables to fit various types of models to the sample data, to evaluate the usefulness of these models,and finally to estimate the average value of y or predict the actual value of y for given values of x1, x2, x3, …, xk . Specific Topics 1. The general linear model and assumptions 2. The method of least squares 3. Analysis of variance for multiple regression 4. Sequential sums of squares 5. The analysis of variance f test 6. The coefficient of determination, R 2 7. Testing the partial regression coefficients 8. Adjusted R 2 9. Residual plots 10. Estimation and prediction using the regression model 11. Polynomial regression model 12. Qualitative variables in a regression model 13. Testing sets of regression coefficients 14. Stepwise regression analysis 15. Causality and multicollinearity 13.1 Introduction Multiple linear regression is an extension of the methodology of simple linear regression to allow for more than one independent variable. For example, a company’s regional sales y of a product might be related to three factors: - x1 — the amount spent on television advertising - x2 — the amount spent on newspaper advertising - x3 — the number of sales representatives assigned to the region. Several questions arise: - How well does the model fit? - How strong is the relationship between y and the predictor variables? - Have any important assumptions been violated? - How good are estimates and predictions? 13.2 The Multiple Regression Model The general linear model for a multiple regression