1 Using Regres-sion to Estimate and Predict

Our discussion of regression will concentrate on linear regression. Its central idea is to find the straight line that "best" fits a set of observed points, i.e., observed pairs of values of two variables. With observed data we often have problems of minor errors of measurement and with the effects of chance deviations from some underlying pattern. For example, height and weight data are subject to errors of measurement (and of reporting!) and are also affected by the fatness or thinness of the persons measured. Nevertheless, there is an underlying pattern of greater weight with greater height and vice versa.