For the column proportions and column means tests, by default columns are tested in groups that are determined by the structure of the table. For example, in a table with three categories in the column variable, columns would be labeled A, B, and C, and the test would be carried out on all combinations of columns A-C. In nested tables, the columns of the inner nest level are repeated (and given unique IDs), and the test is carried out for each set of columns within the nesting level. Alternatively, if columns are being tested only within a variable or nest, you can reuse column IDs by using the ‘Allocate column IDs per variable (test only between the columns in the variable)’ option.

You can select the columns you want to test against each other using the Column Test button in the Statistics tab of the Table Properties dialog.

All combinations of the columns that you specify are tested. Separate the column IDs using a forward slash character, for example:

This tests all combinations of columns A, B, and C.

You can specify separate groupings of columns to test by separating groups of columns with a comma, for example:

This tests all combinations of columns A, C, and E, and all combinations of columns B, D, and F.

Columns that you do not include are omitted from the tests. If you do not specify columns to test, the default column groupings are used.

If you specify columns to test, ensure that the column IDs you specify exist in the table, and that they correspond to valid combinations of columns to test. You can check this by running the test first using the same table specification but with the default column selection.

You can define column IDs as a string using the ColumnsID property. Each character in the string is used to allocate a column ID, with a period or space character used to indicate that an ID should not be allocated. A character needs to be added to the ID's string for each column, even if the column is hidden. When allocating column IDs a character should also be added for the base columns.

When using the “Allocate column IDs per variable” option (Statistics.AllocateColumnIDsPerVariable), the column IDs for each variable can be defined separated by semicolons.

For the following table, the column IDs could be set as follows to test Yes-Male against No-Male, and Yes-Female against No-Female:

Column IDs are only applied to visible table elements. Hidden elements are not taken in consideration when defining a Column IDs as a string.