PROVA.

Can an average improve when no group improves?

Keep two groups’ outcomes fixed and change who makes up the total.

This is a prepared example with fictional records.

Two groups in one program

Keep the group results fixed.

In this illustration, 80% of Group A and 40% of Group B meet the same outcome. The groups differ in circumstances relevant to the outcome. There are 100 people in total.

The same outcome and group rates are used in both mixes.
GroupOriginal countNew countOutcome rate
A202080%, unchanged
B808040%, unchanged
48%overall, compared with 48% in the original mix

(20 × 80% + 80 × 40%) ÷ 100 = 48%