Data Analysis with Microsoft Excel: Updated for Office 2007

Chapter 10 Analysis of Variance 419

plot would seem to indicate that there is no interaction between cola brand

and cola type. To confi rm our visual impression, we’ll have to perform a

two-way analysis of variance.

Using Excel to Perform a Two-Way Analysis of Variance

The Analysis ToolPak provides two versions of the two-way analysis of vari-

ance. One is for situations in which there is no replication of each com-

bination of factor levels. That would be the case in this example if the

experimenter had tested only one can of soda for each cola brand and type.

However, the experiment has been done with six cans, so you should per-

form a two-way analysis of variance with replication.

Note that the number of cans for each cell of brand and type must be the

same. Specifi cally, you cannot use data that have fi ve cans of diet Coke and

six cans of regular Coke. Data with the same number of replications per cell

are called balanced data. If the number of replicates is different for different

combinations of brand and type, you cannot use the Analysis ToolPak’s two-

way analysis of variance command.

Finally, to use the Analysis ToolPak on this data set, it must be organized

in a two-way table. Figure 10-19 shows this table for the cola data. The data

are formatted so that the fi rst factor (the four cola brands) is displayed in the

columns, and the second factor (diet or regular) is shown in the rows of the

table. Replications (the six cans in each pack) occupy six successive rows.

Each cell in the two-way table is the value of the foam volume for a particu-

lar can. You can create this table using the Create Two-Way Table command

included with StatPlus.

Figure 10-19

Two-way

table of

foam values