In the paper "Calibrating Noise to Sensitivity in Private Data Analysis" by Dwork et al., the term "statistical difference" is used as following (in page 280):

Finally, if a $1 − \gamma$ fraction of the $z$’s are $\delta$-good for a particular pair $(r, b)$, then the statistical difference between the distribution $p'(z)$ and $p(z)$ is at most $2(\gamma + \delta)$.

where $\delta$-good is defined as:

A value $z$ is $\delta$-good for a pair $(r, b)$ if $p'(z) − p(z) \leq \delta · p(z)$.

What is the term "statistical difference" as used in this paper? It seems it is not defined anywhere else in the paper and it is used as if something well-known.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.