In this paper, we develop new test statistics for private hypothesis testing.
These statistics are designed specifically so that their asymptotic
distributions, after accounting for noise added for privacy concerns, match the
asymptotics of the classical (non-private) chi-square tests for testing if the
multinomial data parameters lie in lower dimensional manifolds (examples
include goodness of fit and independence testing). Empirically, these new test
statistics outperform prior work, which focused on noisy versions of existing
statistics.
Beschreibung
[1610.07662] A New Class of Private Chi-Square Tests
%0 Journal Article
%1 kifer2016class
%A Kifer, Daniel
%A Rogers, Ryan
%D 2016
%K differential-privacy stats
%T A New Class of Private Chi-Square Tests
%U http://arxiv.org/abs/1610.07662
%X In this paper, we develop new test statistics for private hypothesis testing.
These statistics are designed specifically so that their asymptotic
distributions, after accounting for noise added for privacy concerns, match the
asymptotics of the classical (non-private) chi-square tests for testing if the
multinomial data parameters lie in lower dimensional manifolds (examples
include goodness of fit and independence testing). Empirically, these new test
statistics outperform prior work, which focused on noisy versions of existing
statistics.
@article{kifer2016class,
abstract = {In this paper, we develop new test statistics for private hypothesis testing.
These statistics are designed specifically so that their asymptotic
distributions, after accounting for noise added for privacy concerns, match the
asymptotics of the classical (non-private) chi-square tests for testing if the
multinomial data parameters lie in lower dimensional manifolds (examples
include goodness of fit and independence testing). Empirically, these new test
statistics outperform prior work, which focused on noisy versions of existing
statistics.},
added-at = {2019-09-19T13:35:56.000+0200},
author = {Kifer, Daniel and Rogers, Ryan},
biburl = {https://www.bibsonomy.org/bibtex/2ab8c74bca4e1dfadffd4eda34f196e93/kirk86},
description = {[1610.07662] A New Class of Private Chi-Square Tests},
interhash = {13e69ee22bda84da301b655aa3a9cd20},
intrahash = {ab8c74bca4e1dfadffd4eda34f196e93},
keywords = {differential-privacy stats},
note = {cite arxiv:1610.07662},
timestamp = {2019-09-19T13:35:56.000+0200},
title = {A New Class of Private Chi-Square Tests},
url = {http://arxiv.org/abs/1610.07662},
year = 2016
}