Chapter 10 Analysis of Variance 417
there is much difference due to cola type (diet or regular)? Because the foam
levels do not appear to differ much between the two types, you can expect
that the test for the type effect in a two-way analysis of variance will not be
signifi cant. However, look at the differences among the four brands of colas.
The foamiest can of Pepsi is below the least foamy can of Coke, so you might
expect that there will be a signifi cant cola effect.
The Interaction Plot
The histograms and boxplots give us an idea of the infl uence of cola type
and cola brand on foam volume. How do we graphically examine the in-
teraction between the two factors? We can do so by creating an interaction
plot, which displays the average foam volume for each combination of fac-
tors. To do this, we take advantage of Excel’s pivot table feature.
To set up the pivot table:
1 Click the Cola Data sheet tab to return to the data.
2 Click the PivotTable button from the Tables group of the Insert tab.
3 Verify that the New Worksheet option button is selected and then
click the OK button.
Excel opens a new worksheet displaying the fi elds from the list in a
PivotTable Field list pane.
4 Drag the Type field from the field list and drop it in the Column
Labels box.
5 Drag the Cola fi eld from the list of fi elds and drop it into the Row
Labels box.
6 Drag the Foam fi eld from the fi eld list and drop it in the Values box.
7 Click Sum of Foam in the Values box and select Value Field Settings
from the pop-up menu.
8 Click Average in the Summarize Value field by List box then
click OK.
9 Click the Grand Totals button from the Layout group on the Design
tab of the PivotTable Tools ribbon and select Off for Rows and
Columns to run off the grand total for the rows and columns of the
PivotTable.
