
MODEL BUILDING 181
residuals against estimated values of the dependent variable. A homosce-
dastic model will display a cloud of dots with no visible pattern, whereas
heteroscedasticity will be characterized by a pattern, such as a funnel
shape, indicating greater error as the dependent increases.
Using the Celebrities.xls worksheet discussed previously, multiple
regression was rerun, using SPSS and requesting a plot of the residuals.
This is shown in Figure 4.44. The Y- axis is the standardized residuals and
the X- axis is the standardized predicted values. With homoscedasticity,
you would expect the spread of the data to be ...