The product difference test is not a separate statistical test in its own right. Instead, it enables you to apply statistical testing (using the column proportions or column means tests) to all combinations of categories in a number of variables. One use for this is to identify those attributes of tested products that show significant differences between products.

The test creates a table specification by breaking down a number of variables, known as difference attributes, added to the side of the table, and creating a separate row for each category in each variable, and for each combination of categories from the variables.

For example, if you use the variables education and biology as difference attributes, the test first creates one row for each category (for simplicity, these examples omit the Not Answered categories, but if you choose to include them the table will also include rows for those categories):

If you request two combinations of difference attributes, it also creates a row for each combination of categories in the two variables to give the following rows:

Having created the side of the table, the test applies the column proportions and/or column means test to the table, using the columns in the variable specified on the top of the table. The test also hides any rows that do not contain significant results, and sorts the rows by significance. The end result is a table that displays a detailed breakdown of significant results by combination of categories.