Desktop User Guides > Reporter > Applying statistical tests > Column means test > The least significant difference
The least significant difference
When you add a column means test to a table, you can optionally display a least significant difference column for each row of the table. This is the variance for all specified columns at once, rather than for the pairs of columns that the column means test uses.
When the column means test is applied to a group of columns (for example, ABC), each pair of columns (AB, AC, BC) is tested separately to see if the mean values are significantly different. The least significant difference calculation uses the data from all columns in the group to calculate the smallest difference in means that would be significant at the requested significance level.
To display the least significant difference column on a table
Do one of the following:
Select Show least significant difference on the Statistics tab of the Table Properties dialog.
Use the following script:
TableDoc.Table.Statistics["ColumnMeans"].Properties["ShowLSD"] = True
The least significant difference calculation is available only for independent samples (that is, for non-overlapping columns).
The least significant difference test is available only for tables with a significance level of 1, 5, 10, 15, 20, 25, 32, or 50.
The least significant difference is available with nesting and concatenating on the top axis. It is not available for tables that use the TestColumns option to carry out non-default column testing.
The characters XXXXX are displayed instead of the result if the table does not meet these conditions.
This test is appropriate only when the groups being tested have similar base sizes. If the base sizes differ by more than 25%, the test is not suitable for the table.
See also
Column means test