Desktop User Guides > Professional > Table scripting > Statistical tests > Column means test > Statistical formula for the Column Means Test
 
Statistical formula for the Column Means Test
The column means test is performed separately for each relevant pair of columns within a row that contains mean values and so the formula is presented in terms of one row and one pair of columns.
Notation
Description
This graphic is described in the surrounding text.
Weighted count of cases contributing to the mean in column i.
This graphic is described in the surrounding text.
Sum of squared weights for column i.
This graphic is described in the surrounding text.
Weighted count of cases contributing to the mean for the overlap.
This graphic is described in the surrounding text.
Sum of squared weights for the overlap.
This graphic is described in the surrounding text.
Weighted sum of the values in column i.
This graphic is described in the surrounding text.
Weighted sum of the squared values in column i.
The mean in each column i is
This graphic is described in the surrounding text.
If the effective base is being used, the effective base in each column i is
This graphic is described in the surrounding text.
Otherwise
This graphic is described in the surrounding text.
The test is not performed if:
w i <= 0
The effective base is being used and q i <= 0
The mean values in the two columns being tested are identical
The sample variance in column i is
This graphic is described in the surrounding text.
If we set
This graphic is described in the surrounding text.
Then the pooled estimate of the population variance is
This graphic is described in the surrounding text.
The t value is
This graphic is described in the surrounding text.
With no overlap, Z and e o are both zero. With overlap, Z is 1.0, except in the case of grids, where it is:
This graphic is described in the surrounding text.
where:
X12 is the weighted sum, for respondents in both columns, of the value in column 1 multiplied by the value in column 2
all X and Y terms in Z refer to respondents who are in both columns.
The degrees of freedom, DF, are
DF = e1 + e2 - e0 - 2
See also
Column means test