344 Statistical Methods
Depending on the number of variables you are plotting, SPLOMs can be
diffi cult to view on the screen. If you can’t see the entire SPLOM on your
screen, consider reducing the value in the Zoom Control box. You can also
reduce the SPLOM by selecting it and dragging one of the resizing handles
to make it smaller.
How should you interpret the SPLOM? Each of the fi ve variables is plot-
ted against the other four variables, with the four plots displayed in a row.
For example, ACT Math is plotted as the y variable against the other four
variables in the fi rst row of the SPLOM. The fi rst plot in the fi rst row is ACT
Math versus Alg Place, and so on. On the other hand, the fi rst plot in the
fi rst column displays Alg Place as the y variable and is plotted against the
x variable, ACT Math. The scales of the plot are not shown in order to save
space. If you fi nd a plot of interest, you can recreate it using Excel’s Chart
Wizard to show more details and information.
Carefully consider the plots in the second to last row, which show Calc
against the other variables. Each plot shows a roughly linear upward trend.
It would be reasonable to conclude here that correlation and linear regres-
sion are appropriate when predicting Calc from ACT Math, Alg2 Grade, Alg
Place, and HS Rank.
Recall from Figure 8-24 that Alg Place had the highest correlation with
Calc. How is that evident here? A good predictor has good accuracy, which
means that the range of y is small for each x. Of the four plots in the fourth
row, the plot of Calc against Alg Place has the narrowest range of y values
for each x. However, Alg Place is the best of a weak lot. None of the plots
shows that really accurate prediction is possible. None of these plots showsFigure 8-25
Scatter plot
matrix