Statistics for Business: Decision Making and Analysis, 3rd Edition

Chapter Summary

Reliable use of the simple regression model requires that we check for several common problems: heteroscedasticity, outliers, and dependence. Heteroscedasticity occurs when the underlying errors lack constant variance. Homoscedasticity occurs when random variables have equal variation. Transformations may correct this problem. Outliers, particularly leveraged outliers at the boundaries of the explanatory variable, can exert a strong pull on the model. Comparisons of the fit with and without an outlier are useful. Dependent residuals indicate that the model errors are dependent, violating a key assumption of the SRM. For time series, the Durbin-Watson statistic tests for a particular type of dependence known as autocorrelation ...