While reviewing some slides I see the following formula for calculating lift (which is defined as "measure of dependent/correlated events"):
$$lift(A,B) = \frac{s(A\cup B)}{s(A)\cdot s(B)}$$
which for this table:
\begin{matrix} & A & \overline{A}\\ B & 400 & 350 \\ \overline{B} & 200 & 50 \end{matrix}
gives the following result: 0.4 / (0.6 * 0.75) = 0.89
and tells that if lift = 1 then events are independent, if < 1 negatively correlated and if > 1 - positively.
I can understand how the calculations are made, but I can not understand the reason behind it and how correlation is connected (wanted to write correlate :-) ) to the number 1.
I have seen a question here about lift, but it looks to me like a completely different question.
