Data

For this exercise, we utilize the ``cars'' data downloaded from the StatLib data repository (see the link on the course projects webpage). There are eight variables:

$y$

mpg: miles per gallon

$x_1$

cylinders: number of cylinders

$x_2$

displacement: engine displacment (in cubic inches?)

$x_3$

horsepower: engine power

$x_4$

weight: in pounds

$x_5$

acceleration: I don't know the units

$x_6$

model.year: 70 to 82

$x_7$ and $x_8$

origin: 1=USA, 2=Europe, 3=Japan

The response variable is mpg; we want to predict fuel efficiency from the other variables. Note that origin is a qualitative variable. We included its effect by allowing a different intercept. This was accomplished by including two ``indicator'' variables $x_7$ and $x_8$. The variable euro $x_7 = 1$ if the car is made in Europe and otherwise $x_7 = 0$. The variable japan $x_8 = 1$ if the car was made in Japan and otherwise $x_8 = 0$.

There were originally 406 observations, but 14 observations had missing values in one or more variables, and they were deleted, leaving $n = 392$ observations. The minitab projects may be downloaded here: cars1.MPJ and cars2.MPJ. The cars2.MPJ project includes the ``validation data'' (described below) in separate columns. Further description of the output is given below.