Some Common Plotting Functions

Functions that create a new picture are called ``high-level'' in S-PLUS terminology. Some useful high-level functions are hist (produces a histogram), qqnorm (produces a normal quantile plot) and plot (described in more detail below).

Let's begin our analysis of the education data by examining the distribution of the response variable, SE70.

> hist(SE70)

> qqnorm(SE70)

From these two plots it appears that school expenditures may be close to normally distributed, except for one outlier which had school expenditures over 350. If you look through the data, you will see that the outlier is Alaska. Let's see the histogram again, without Alaska. Alaska is listed 50th in the file, so we want to plot SE70[-50] (school expenditures of every state but the 50th).

> hist(SE70[-50])

Judging by this histogram, SE70 does not have a particularly normal distribution after all. Of course, you should check this by doing another normal quantile plot.

The next step is to examine the relationship between SE70 and the other variables. The function plot will come in handy here. This function can be called in several different ways, and since S-PLUS uses Object Oriented Programming, plot will produce a different sort of plot depending on the class of object it receives. The most common use is plot(x,y), with x a numeric vector, which results in a scatterplot of x against y. If x is a factor, this will instead produce boxplots.

We can expect that school expenditures will be related to the number of students, so let's plot Y69 (students) against SE70 .

> plot(Y69, SE70)

It does not appear from this scatterplot that there is any particular relationship between number of students and school expenditures (note that Alaska is the point in the upper right-hand corner).

We can also plot the Region vector (a factor) against school expenditures.

This shows that school expenditures are generally lower in the South.

The function plot can also be called with a single argument, and the output will be a plot of an index against that vector (if the argument is a vector) or a series of diagnostic plots (if the argument is a model, see the handout on models for an example).

Question: How do you do the above boxplots, excluding Alaska? How do you plot only the North East and North Central data?