Differentially Private Joint Independence Test

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Xingwei, Chen, Yuexin, Xu, Wangli
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908307997851648
author Liu, Xingwei
Chen, Yuexin
Xu, Wangli
author_facet Liu, Xingwei
Chen, Yuexin
Xu, Wangli
contents Identification of joint dependence among more than two random vectors plays an important role in many statistical applications, where the data may contain sensitive or confidential information. In this paper, we consider the the $d$-variable Hilbert-Schmidt independence criterion (dHSIC) in the context of differential privacy. Given the limiting distribution of the empirical estimate of dHSIC is complicated Gaussian chaos, constructing tests in the non-privacy regime is typically based on permutation and bootstrap. To detect joint dependence in privacy, we propose a dHSIC-based testing procedure by employing a differentially private permutation methodology. Our method enjoys privacy guarantee, valid level and pointwise consistency, while the bootstrap counterpart suffers inconsistent power. We further investigate the uniform power of the proposed test in dHSIC metric and $L_2$ metric, indicating that the proposed test attains the minimax optimal power across different privacy regimes. As a byproduct, our results also contain the pointwise and uniform power of the non-private permutation dHSIC, addressing an unsolved question remained in Pfister et al. (2018). Both numerical simulations and real data analysis on causal inference suggest our proposed test performs well empirically.
format Preprint
id arxiv_https___arxiv_org_abs_2503_18721
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Differentially Private Joint Independence Test
Liu, Xingwei
Chen, Yuexin
Xu, Wangli
Statistics Theory
Cryptography and Security
Machine Learning
Methodology
62G10, 62H20
Identification of joint dependence among more than two random vectors plays an important role in many statistical applications, where the data may contain sensitive or confidential information. In this paper, we consider the the $d$-variable Hilbert-Schmidt independence criterion (dHSIC) in the context of differential privacy. Given the limiting distribution of the empirical estimate of dHSIC is complicated Gaussian chaos, constructing tests in the non-privacy regime is typically based on permutation and bootstrap. To detect joint dependence in privacy, we propose a dHSIC-based testing procedure by employing a differentially private permutation methodology. Our method enjoys privacy guarantee, valid level and pointwise consistency, while the bootstrap counterpart suffers inconsistent power. We further investigate the uniform power of the proposed test in dHSIC metric and $L_2$ metric, indicating that the proposed test attains the minimax optimal power across different privacy regimes. As a byproduct, our results also contain the pointwise and uniform power of the non-private permutation dHSIC, addressing an unsolved question remained in Pfister et al. (2018). Both numerical simulations and real data analysis on causal inference suggest our proposed test performs well empirically.
title Differentially Private Joint Independence Test
topic Statistics Theory
Cryptography and Security
Machine Learning
Methodology
62G10, 62H20
url https://arxiv.org/abs/2503.18721