Adding a product difference test

You can add a product difference test to a table containing a column proportions or a column means test, to break down the results by category using a number of different variables, known as difference attributes. This enables you to see which combinations of categories produce significant results.

1 Create a table with either no variables or one variable on the side and one variable on the top.

You can use an edited variable provided that you edit the variable before you specify the product difference test. Once you have specified a product difference test for a table, the entire table specification is frozen and any further edits that you make to variables will not be reflected in the table.

2 In the Variables pane, use Shift+click or Ctrl+click to select the variables you want to use as the difference attributes.

3 Click Tables > Add Difference Attributes.

4 Choose the number of combinations to use:

If you select one combination, each category in each of the difference attributes will be combined with each category in the variable on the side of the table to form a separate row of the table. Each row will be tested for significance. Each row therefore contains a combination of two categories.

If you select two combinations, in addition to the rows produced by the above set of combinations, a row is created for each combination among the two difference attributes and the side variable. Each of these rows contains a combination of three categories.

If you select three combinations, in addition to the 1-and 2-combination rows, a row is created for each combination of categories in all combinations of three difference attributes and the side variable. Each of these rows contains a combination of four categories.

Depending on the numbers of categories in each of your variables, this can result in a very large table, which might take a long time to generate and which might not contain any significant results unless your survey data contains a sufficiently large set of responses.

5 By default, the column proportions test will be applied to the table. If you want to change this, clear Apply default statistics (column proportions).

6 By default, only significant results are displayed in the table. If you want to see all results, clear Apply default hide rules.

On the Design pane, the table specification is now frozen, and the size of the table appears as Difference_Attributes_axis, representing the combinations of categories from the side variable and the difference attributes that you selected.

8 If you deselected the default statistics check box, you must also select the statistical test you want to include using the Table Properties dialog. Click Tables > Properties, and then select the appropriate options in the Statistics tab. Select the Difference Attributes dialog again, and then click OK.

If you do not see any results, this might be because your survey data contains insufficient records to support the number of attributes you have selected. Alternatively, it might be that none of the results is significant. Try selecting fewer attributes, or setting a lower significance level.