Searcharxiv⌕ Search

arXiv subjects

Xianyang Zhang

Publications and source records attributed to Xianyang Zhang.

At least 19 recordsLinked to original sources

Testing Conditional Mean Independence

This article introduces a novel nonparametric test for conditional mean independence, which formally evaluates whether a random vector X has an additional impact on the mean regression function of a response Y when a conditioning random vector Z is already included. By framing the problem as classical partial effect testing, our approach goes beyond traditional parametric models to provide a highly general and flexible framework. The test statistic is constructed using martingale difference divergence (MDD) after neutralizing the effect of Z on Y via a sieve approximation. The population residual MDD characterizes conditional mean independence under suitable moment conditions. We present a theoretical framework for the test under both a fixed number of basis functions and a growing sieve dimension. We derive the nonstandard and non-pivotal asymptotic null distributions in both regimes. To facilitate practical implementation, we propose a wild bootstrap procedure to approximate these limiting distributions and establish its asymptotic consistency. To establish the fundamental properties required for the test to be widely useful, we conduct a power analysis, demonstrating the test's consistency under fixed alternatives and establishing its asymptotic power against local alternatives. The practical utility and finite-sample performance of the proposed test are examined through simulation studies and an exploratory application to a real-world variable selection problem in regression modeling. The code for the proposed test and numerical simulations can be found at: https://github.com/leece087/Conditional-Mean-Independence-Test.

stat.ME↗

U-centering as subset ANOVA: edge regression and higher-order theory

Unbiased squared distance covariance and the Hilbert-Schmidt independence criterion (HSIC) are fourth-order U-statistics, yet U-centering evaluates them from pairwise arrays in $O(n^2)$ operations. We show that U-centering is exactly least-squares residualization after fitting additive endpoint effects. This analysis of variance (ANOVA) interpretation explains the zero row sums and the normalization through residual degrees of freedom. The same geometry gives regression representations of dependence and two-sample statistics, exact permutation variances, and relations among pooled and regression mean squares. For arrays indexed by $r$-element subsets and $n\ge2r$, higher-order U-centering removes all lower-order subset effects and projects onto a space of dimension $\binom{n}{r}-\binom{n}{r-1}$. For fixed symmetric square-integrable kernels evaluated on independent and identically distributed (i.i.d.) paired observations, we derive the residual covariance matrix, exact mean-squared errors for estimating the highest-order Hoeffding components, unbiased positive semidefinite (PSD) covariance estimates, and moment formulas under independence and permutation. The normalized residual inner product equals a direct unbiased estimator involving up to $2r$ observations, but can be evaluated in $O(n^r)$ operations for fixed $r$. For four-argument matching kernels, three matrix products yield exactly unbiased estimators of a fourth spectral moment of kernel cross-covariance in $O(n^3)$ arithmetic and $O(n^2)$ storage. A numerical illustration distinguishes dependence structures with equal HSIC and different spectral concentration.

math.ST↗

Exact Simes inequalities for two-sided Gaussian tests

The Simes test combines marginal p-values to test a global null hypothesis. Whether it controls the type I error for two-sided Gaussian tests with arbitrary covariance remains open. We prove that, for every fixed number $m$ of tests, the rejection probability does not exceed the nominal level when that level is sufficiently small. The same range of levels works for every covariance matrix. Equality holds if and only if the p-values form independent groups with identical values within each group. For $m=3$, a computer-assisted proof establishes validity at $0.05$, and an analytic argument extends it to an explicit interval around $0.05$. As the number $m$ of tests increases, we prove validity in two settings: each test statistic is correlated with at most a fixed number of others; or the statistics form independent blocks whose sizes grow sufficiently slowly, at a specified rate of order $\log\log m$. Both settings allow arbitrary correlations within these restrictions. For any fixed $0<a<1$, validity holds for every $m$ above a threshold $M$ and every level $0<α\le a$; the same $M$ works for all allowed covariances. Further results quantify the difference between the nominal level and the rejection probability, and permit more correlated neighbors when their correlations are bounded away from $\pm1$. We do not obtain numerical values for the general small-level cutoffs or the required number of tests.

math.ST↗

Fair Regression under Approximate Demographic Parity

Exact demographic parity requires the same prediction distribution across sensitive groups. In regression, this can be too restrictive and reduce predictive accuracy. We instead bound every pairwise Wasserstein distance between group-specific prediction distributions by a chosen tolerance, allowing the groups to differ while keeping fairness tunable. On the real line, the prediction and constrained transport problems have the same infimum. If the transport problem attains this value and suitable Monge maps exist, a deterministic transport-form predictor realizes it. Under a standard one-dimensional condition, the problem becomes a bounded quantile optimization that is convex when the conditional expected loss is convex in prediction. Under squared loss with two positively weighted groups, the $W_1$, $W_2$, and $W_\infty$ constraints yield soft-thresholding, shrinkage, and clipping. Under the stated conditions, we establish rates for the known-cost plug-in estimator. When the cost is unavailable, we describe an observed-loss computational proxy. Synthetic and California 2018 ACS studies illustrate the tunable fairness-accuracy trade-off and distinguish calibration feasibility from held-out fairness and accuracy.

stat.ML↗

Mirror and knockoff+ thresholds under dependence

Many multiple-testing procedures control the false discovery rate (FDR) by comparing the two tails of a null distribution. At a fixed cutoff, marginal symmetry makes this natural. Mirror and knockoff+ thresholds select the cutoff from the same data, so the standard finite-sample guarantee uses a stronger property: conditional on magnitudes and nonnull scores, null signs are independent fair coins. Failure can be severe without this property. Models satisfying positive regression dependence on a subset (PRDS) can have exactly uniform null $p$-values and large FDR. Under every fixed positive Gaussian equicorrelation, the all-null FDR converges to one half. Opposing loadings in Gaussian factor models can make FDR and power arbitrarily close to one; near-total failure also occurs for exchangeable, pairwise-uncorrelated scores. At a nominal input level $q<1/2$, no deterministic rule based only on the two tail counts can both reject and control FDR uniformly over our class if more discoveries or fewer controls cannot make rejection harder. We give finite-sample repairs based on joint sign information. Conditional sign odds may be bounded outside an exceptional event or averaged over negative controls; neither route uniformly dominates, and the integrated bounds are sharp. Independent calibration data or a specified Gaussian joint model yield valid adjusted levels. Simulations show that integration retains more power under diffuse Gaussian dependence, whereas exceptional-event calibration is more powerful when very large odds occur only for rare aligned signs; unadjusted FDR exceeds the target in both settings. Covariance alone is insufficient outside a specified joint model. Thus the relevant boundary is not marginal symmetry but joint information that remains valid after adaptive cutoff selection.

math.ST↗

High-dimensional Change-point Detection Using Generalized Homogeneity Metrics

Change-point detection is a classical problem in statistics. We address the problem of detecting abrupt changes in the data-generating distributions of a sequence of high-dimensional observations beyond the first two moments. This problem remains less explored, especially in the high-dimensional context, compared to detecting changes in the mean or the covariance structure. To the best of our knowledge, this is one of the first attempts to detect and localize general types of distributional changes in the high-dimensional regime. We develop a distance-based method to (i) test for the existence of a change-point, and (ii) identify the change-point locations in an independent sequence of high-dimensional observations. Our approach rests upon recent distance-based tests for the homogeneity of two high-dimensional distributions. We construct a single change-point test statistic based on a cumulative sum process in an embedded Hilbert space and rigorously derive its limiting null distribution and prove asymptotic consistency under the high-dimensional medium sample size (HDMSS) framework. Subsequently, we combine our statistics with the Narrowest-Over-Threshold (NOT) strategy to recursively estimate and test for multiple change-point locations. We also study a componentwise monotone-invariant, rank-based extension; because its pseudo-observations are pooled empirical mid-ranks and are therefore dependent, we present this version as a practically useful heuristic extension supported by simulation evidence rather than as a fully proved analogue of the original statistic. The superior performance of our methodology compared to existing procedures is illustrated via extensive simulation studies and an application to U.S. stock return data during the global financial crisis. The proposed method is implemented in the R package KDist, available at https://github.com/zhangxiany-tamu/KDist.

stat.ME↗

Chernoff's Density Is Strongly Log-Concave

Let $f$ be the density of the Chernoff random variable $\mathrm{argmax}_{t\in\mathbb{R}}\{W(t)-t^2\}$, where $W$ is a two-sided Brownian motion. This note proves the conjecture of Balabdaoui and Wellner (2014) that $f$ is strongly log-concave. The proof was generated in its entirety by GPT-5.6 Sol.

math.PR↗

Testing composite null hypotheses with high-dimensional dependent data: a computationally scalable FDR-controlling procedure

Testing composite null hypotheses is fundamental to many scientific applications, including mediation and replicability analyses, and becomes particularly challenging in high-throughput settings involving tens of thousands of features. Existing high-dimensional composite null hypotheses testing often ignores the dependence structure among features, leading to overly conservative or liberal results. To address this limitation, we develop a four-state hidden Markov model (HMM) for bivariate $p$-value sequences arising from two-study replicability analysis. This model captures local dependence among features and accommodates study-specific heterogeneity. Based on the HMM, we propose a multiple testing procedure that asymptotically controls the false discovery rate (FDR). Extending this framework to more than two studies is computationally intensive, with complexity growing exponentially in the number of studies $n$. To address this scalability issue, we introduce a novel e-value framework that reduces computational complexity to quadratic in $n$, while preserving asymptotic FDR control. Extensive simulations demonstrate that our method achieves higher power than existing approaches at the same FDR levels. When applied to genome-wide association studies (GWAS), the proposed approach identifies replicable SNP-level signals that are not detected at the same significance threshold by competing methods.

stat.ME↗

Statistical Validation of Computer Models: Global and Subdomain Hypothesis Testing

Computer simulations play an important role in scientific discovery and engineering innovation. Reliable computer models enable virtual experimentation that reduces the need for costly and time-consuming physical testing. However, the credibility of such models hinges on rigorous statistical validation against real-world data. This paper develops a formal frequentist framework for both global and subdomain validation of computer models. We propose the Fourier Maximum Modulus Test (FMMT), which leverages kernel ridge regression (KRR) to estimate the discrepancy between the computer model and the physical process, followed by a frequency-domain test based on weighted generalized Fourier coefficients. The theoretical analysis establishes the asymptotic normality of these coefficients, allowing for closed-form p-values. Simulation studies and a shear-layer experiment demonstrate that FMMT achieves high power, accurate Type I error control, and strong sensitivity to localized discrepancies.

stat.ME↗

A Modern Theory for High-dimensional Cox Regression Models

The proportional hazards model has been extensively used in many fields such as biomedicine to estimate and perform statistical significance testing on the effects of covariates influencing the survival time of patients. The classical theory of maximum partial-likelihood estimation (MPLE) is used by most software packages to produce inference, e.g., the coxph function in R and the PHREG procedure in SAS. In this paper, we investigate the asymptotic behavior of the MPLE in the regime in which the number of parameters p is of the same order as the number of samples n. The main results are (i) existence of the MPLE undergoes a sharp 'phase transition'; (ii) the classical MPLE theory leads to invalid inference in the high-dimensional regime. We show that the asymptotic behavior of the MPLE is governed by a new asymptotic theory. These findings are further corroborated through numerical studies. The main technical tool in our proofs is the Convex Gaussian Min-max Theorem (CGMT), which has not been previously used in the analysis of partial likelihood. Our results thus extend the scope of CGMT and shed new light on the use of CGMT for examining the existence of MPLE and non-separable objective functions.

math.ST↗

Fair Regression under Demographic Parity: A Unified Framework

We propose a unified framework for fair regression tasks formulated as risk minimization problems subject to a demographic parity constraint. Unlike many existing approaches that are limited to specific loss functions or rely on challenging non-convex optimization, our framework is applicable to a broad spectrum of regression tasks. Examples include linear regression with squared loss, binary classification with cross-entropy loss, quantile regression with pinball loss, and robust regression with Huber loss. We derive a novel characterization of the fair risk minimizer, which yields a computationally efficient estimation procedure for general loss functions. Theoretically, we establish the asymptotic consistency of the proposed estimator and derive its convergence rates under mild assumptions. We illustrate the method's versatility through detailed discussions of several common loss functions. Numerical results demonstrate that our approach effectively minimizes risk while satisfying fairness constraints across various regression settings.

stat.ME↗

Powerful Large-scale Inference in High Dimensional Mediation Analysis

In genome-wide epigenetic studies, exposures (e.g., Single Nucleotide Polymorphisms) affect outcomes (e.g., gene expression) through intermediate variables such as DNA methylation. Mediation analysis offers a way to study these intermediate variables and identify the presence or absence of causal mediation effects. Testing for mediation effects lead to a composite null hypothesis. Existing methods like the Sobel's test or the Max-P test are often underpowered because 1) statistical inference is often conducted based on distributions determined under a subset of the null and 2) they are not designed to shoulder the multiple testing burden. To tackle these issues, we introduce a technique called MLFDR (Mediation Analysis using Local False Discovery Rates) for high dimensional mediation analysis, which uses the local False Discovery Rates based on the coefficients of the structural equation model specifying the mediation relationship to construct a rejection region. We have shown theoretically as well as through simulation studies that in the high-dimensional setting, the new method of identifying the mediating variables controls the FDR asymptotically and performs better with respect to power than several existing methods such as DACT (Liu et al.)and JS-mixture (Dai et al).

stat.ME↗

Nonparametric Block Bootstrap Kolmogorov-Smirnov Goodness-of-Fit Test

The Kolmogorov--Smirnov (KS) test is a widely used statistical test that assesses the conformity of a sample to a specified distribution. Its efficacy, however, diminishes with serially dependent data and when parameters within the hypothesized distribution are unknown. For independent data, parametric and nonparametric bootstrap procedures are available to adjust for estimated parameters. For serially dependent stationary data, parametric bootstrap has been developed with a working serial dependence structure. A counterpart for the nonparametric bootstrap approach, which needs a bias correction, has not been studied. Addressing this gap, our study introduces a bias correction method employing a nonparametric block bootstrap, which approximates the distribution of the KS statistic in assessing the goodness-of-fit of the marginal distribution of a stationary series, accounting for unspecified serial dependence and unspecified parameters. We assess its effectiveness through simulations, scrutinizing both its size and power. The practicality of our method is further illustrated with an examination of stock returns from the S\&P 500 index, showcasing its utility in real-world applications.

stat.ME↗

Distance and Kernel-Based Measures for Global and Local Two-Sample Conditional Distribution Testing

Testing the equality of two conditional distributions is crucial in various modern applications, including transfer learning and causal inference. Despite its importance, this fundamental problem has received surprisingly little attention in the literature, with existing works focusing exclusively on global two-sample conditional distribution testing. Based on distance and kernel methods, this paper presents the first unified framework for both global and local two-sample conditional distribution testing. To this end, we introduce distance and kernel-based measures that characterize the homogeneity of two conditional distributions. Drawing from the concept of conditional U-statistics, we propose consistent estimators for these measures. Theoretically, we derive the convergence rates and the asymptotic distributions of the estimators under both the null and alternative hypotheses. Utilizing these measures, along with a local bootstrap approach, we develop global and local tests that can detect discrepancies between two conditional distributions at global and local levels, respectively. Our tests demonstrate reliable performance through simulations and real data analysis.

stat.ME↗

A General Framework for Multiple Testing via E-value Aggregation and Data-Dependent Weighting

Motivated by recent findings in Li and Zhang (2025), which established an equivalence between certain p-value-based multiple testing procedures and the e-Benjamini-Hochberg procedure (Wang and Ramdas, 2022), we introduce a general framework for constructing novel multiple testing methods through the aggregation and combination of e-values. Specifically, we propose methodologies for three distinct scenarios: (i) assembly of e-values obtained from different subsets of data, simultaneously controlling group-wise and overall false discovery rates; (ii) aggregation of e-values derived from different procedures or the same procedure employing different test statistics; and (iii) adaptive multiple testing methods that incorporate external structural information to enhance statistical power. A notable feature of our approach is the use of data-dependent weighting of e-values, significantly improving the efficiency of the resulting e-Benjamini-Hochberg procedures. The construction of these weights is non-trivial and inspired by leave-one-out analysis, a widely utilized technique for proving false discovery rate control in p-value-based methodologies. We theoretically establish that the proposed e-Benjamini-Hochberg procedures, when equipped with data-dependent weights, guarantee finite-sample false discovery rate control across all three considered applications. Additionally, numerical studies illustrate the efficacy and advantages of the proposed methods within each application scenario.

stat.ME↗

Generalization Bounds and Model Complexity for Kolmogorov-Arnold Networks

Kolmogorov-Arnold Network (KAN) is a network structure recently proposed by Liu et al. (2024) that offers improved interpretability and a more parsimonious design in many science-oriented tasks compared to multi-layer perceptrons. This work provides a rigorous theoretical analysis of KAN by establishing generalization bounds for KAN equipped with activation functions that are either represented by linear combinations of basis functions or lying in a low-rank Reproducing Kernel Hilbert Space (RKHS). In the first case, the generalization bound accommodates various choices of basis functions in forming the activation functions in each layer of KAN and is adapted to different operator norms at each layer. For a particular choice of operator norms, the bound scales with the $l_1$ norm of the coefficient matrices and the Lipschitz constants for the activation functions, and it has no dependence on combinatorial parameters (e.g., number of nodes) outside of logarithmic factors. Moreover, our result does not require the boundedness assumption on the loss function and, hence, is applicable to a general class of regression-type loss functions. In the low-rank case, the generalization bound scales polynomially with the underlying ranks as well as the Lipschitz constants of the activation functions in each layer. These bounds are empirically investigated for KANs trained with stochastic gradient descent on simulated and real data sets. The numerical results demonstrate the practical relevance of these bounds.

cs.LG↗

Subsampling-based Tests in Mediation Analysis

Testing for mediation effect poses a challenge since the null hypothesis (i.e., the absence of mediation effects) is composite, making most existing mediation tests quite conservative and often underpowered. In this work, we propose a subsampling-based procedure to construct a test statistic whose asymptotic null distribution is pivotal and remains the same regardless of the three null cases encountered in mediation analysis. The method, when combined with the popular Sobel test, leads to an accurate size control under the null. We further introduce a Cauchy combination test to construct p-values from different subsample splits, which reduces variability in the testing results and increases detection power. Through numerical studies, our approach has demonstrated a more accurate size and higher detection power than the competing classical and contemporary methods.

stat.ME↗

Kernel Two-Sample Tests in High Dimension: Interplay Between Moment Discrepancy and Dimension-and-Sample Orders

Motivated by the increasing use of kernel-based metrics for high-dimensional and large-scale data, we study the asymptotic behavior of kernel two-sample tests when the dimension and sample sizes both diverge to infinity. We focus on the maximum mean discrepancy (MMD) using isotropic kernel, including MMD with the Gaussian kernel and the Laplace kernel, and the energy distance as special cases. We derive asymptotic expansions of the kernel two-sample statistics, based on which we establish the central limit theorem (CLT) under both the null hypothesis and the local and fixed alternatives. The new non-null CLT results allow us to perform asymptotic exact power analysis, which reveals a delicate interplay between the moment discrepancy that can be detected by the kernel two-sample tests and the dimension-and-sample orders. The asymptotic theory is further corroborated through numerical studies.

math.ST↗