CHAPTER 2

Apuntes

CHAPTER 2 Checking the Straight Line Fit We discuss basic methods of checking a fitted regression model. Although we talk about these in terms of fitting a straight line, the basic methods apply generally whenever a linear model is fitted, no matter how many predictors there are. Other techniques too advanced for our current context are given in Chapter 8. Here we examine the following: 1. The lack of fit F-test when the data contain repeat observations, that is, when pure error is available (Sections 2.1 and 2.2). 2. Basic visual checks that can be made on the residuals ei = Yi - Yi (Sections 2.3-2.6). 3. The Durbin-Watson test for checking serial correlation (Section 2.7). 2.1. LACK OF FIT AND PURE ERROR General Discussion of Variance and Bias We have already remarked that the fitted regression line is a calculated line based on a certain model or assumption, an assumption we should not blindly accept but should tentatively entertain. In certain circumstances we can check whether or not the model is correct. First, we can examine the consequences of an incorrect model. Let us r...