1: \begin{definition}[Statistical parity---\sp]

2: A predictor $f$ satisfies statistical parity under a distribution over $(X,A,Y)$ if $f(X)$ is independent of the protected attribute $A$. Since $f(X)\in[0,1]$, this is equivalent to $\P[f(X)\ge z\given A=a]=\P[f(X)\ge z]$ for all $a\in\calA$ and $z\in[0,1]$.%

3: \footnote{%

4: A standard definition of statistical independence requires that $\P[f(X)\in S\given A=a]=\P[f(X)\in S]$ for all measurable sets $S$.

5: Since $f(X)$ is a real-valued random variable under Borel $\sigma$-algebra, it is fully characterized by its cumulative distribution function, and

6: so it suffices to consider sets $S=[0,z]$ for $z\in[0,1]$ (see, e.g., Theorem~10.49 of \mbox{\citealp{aliprantis2006infinite}}).}

7: \label{defn:dp}

8: \end{definition}

9: