4 Gaussian linear model

This section covers confidence and prediction intervals, diagnostic plots and quantile-quantile plots.

We present a worked-out example of a linear model fit to the mtcars data set

The data was extracted from the 1974 Motor Trend US magazine, and comprises fuel consumption and 10 aspects of automobile design and performance for 32 automobiles (1973–74 models).

Residual plots are useful to diagnostic

• Misspecification of the response surface (nonlinearity, omitted variables)
• heteroscedasticity
• outliers
• autocorrelation (lack of independence of error terms) if observations are time ordered
• normality assumption.