Departure from overall distribution

Departure from overall distribution is an effect size measurement for the distribution test. It measures the extent to which a conditional distribution does not match the marginal distribution of the target field.

A large effect size suggests that the conditional distribution is very different from the overall target distribution.

The following procedure describes how the effect size is calculated:

  1. Subtract 1 from the number of target categories.
  2. Multiply this result by the total count in the conditional group.
  3. Divide the chi-square value of the distribution test by this result.
  4. Take the square root.
Table 1. Interpretation of effect size
Effect size (ES) Interpretation
ES ≤ 0.2 The result is weak. Although the result is statistically significant, the conditional distribution does not differ much from the overall target distribution.
0.2 < ES ≤ 0.6 The result is moderate. The conditional distribution differs a moderate amount from the overall target distribution.
ES > 0.6 The result is strong. The conditional distribution is very different from the overall target distribution.