Data Analysis with Microsoft Excel: Updated for Office 2007

Chapter 10 Analysis of Variance 415

Because of the number of charts, you must either reduce the zoom factor

on your worksheet or scroll vertically through the worksheet to see all the

plots. Do you see major differences in spread among the eight groups? If so,

it would suggest a violation of the equal-variances assumption, because all

of the groups are supposed to have the same population variance. The his-

tograms seem to indicate a greater variability in the generic colas and the

Shasta brand, whereas less variability is indicated for the Coke and Pepsi

brands. Once again, the two-way ANOVA is fairly robust with respect to the

constant variance assumption, so this might not invalidate the analysis.

You should also look for outliers because extreme observations can make

a big difference in the results. An outlier could be the result of a strange can

of cola, a wrong observation, a recording error, or an error in entering the

data. To gain further insight into the distribution of the data, create a box-

plot of each of the eight combinations of brand and type.

To create boxplots of the foam data:

1 Click Single Variable Charts from the StatPlus menu and then click

Boxplots.

2 Click the Data Values button, select Foam from the range names list,

and click OK.

Figure 10-16
Multiple
histograms
of cola type