arXiv subjects

Yinqiu He

Publications and source records attributed to Yinqiu He.

15 recordsLinked to original sources

arXiv2026

LLM Evaluation on Unseen Questions: Contextual Multidimensional IRT Model

Evaluation of large language models (LLMs) increasingly requires predicting how a model will perform on new questions or tasks before collecting large amounts of new annotations. This problem is challenging because question difficulty, scenario, and underlying capability demands can vary substantially. Simple retrospective averages may confound model ability with item characteristics. In this paper, we study a model-based evaluation framework that combines multidimensional item response theory model with question contexts to predict LLM performance on unseen questions. The framework represents LLMs through latent capability profiles while using question content to inform item characteristics, allowing information to transfer beyond previously observed items. Empirically, we find that for within-scenario evaluation, incorporating question embeddings improves prediction relative to model-free baselines, and that multidimensional latent structure provides a richer description of capability variation than unidimensional alternatives. At the same time, our results reveal an important limitation that the generalizability does not necessarily translate into reliable prediction under cross-scenario shift. These findings suggest that context-aware psychometric modeling is a promising direction for efficient and interpretable LLM evaluation, while also highlighting cross-scenario generalization as a central open challenge.

cs.CL↗

arXiv2026

Bridging Theory and Practice: Statistical Inference for Latent Space Models of Networks

Latent space models have been widely adopted in modeling network data. Developing statistical inference for estimated model parameters enables quantifying associated uncertainty and is pivotal for downstream tasks. Despite recent progress on statistical inference of maximum likelihood estimation, crucial gaps remain between asymptotic theoretical guarantees and practical use. Specifically, how are the oracle maximum likelihood estimators related to the solutions produced by algorithms in practice? Can rigorous guarantees be established for existing algorithms without unnecessary restrictions? To address these fundamental questions, we develop a unified analytical framework that bridges theory and practice of statistical inference for latent space models. First, for the maximum likelihood estimation, we relax the spectral-multiplicity constraint in the existing asymptotic theory to broaden the applicability. Second, we overcome the dependence on unknown true parameters in prior algorithmic analyses by developing novel adaptive criteria and theoretical tools. For the widely used algorithm based on the projected gradient descent and the singular value thresholding, we explicitly connect their outputs to the maximum likelihood estimator without relying on unknown information. Our results provide a solid foundation for practically useful and statistically principled statistical inference in network analysis.

math.ST↗

arXiv2026

Boosting multi-view association testing via devariation

Understanding the interplay between high-dimensional data from different views is essential in biomedical research, particularly in fields such as genomics, neuroimaging and biobank-scale studies involving high-dimensional features. Existing statistical tests for the association between two random vectors often do not fully capture dependencies between views due to limitations in modeling within-view dependencies, particularly in high-dimensional data without clear dependency patterns, which can lead to a potential loss of statistical power. In this work, we propose a novel approach termed devariation which is considered a simple yet effective preprocessing method to address the limitations by adopting a penalized low-rank factor model to flexibly capture within-view dependencies. Theoretical analysis of asymptotic power shows that devariation increases statistical power, especially when within-view correlations impact signal-to-noise ratios, while maintaining robustness in scenarios without strong internal correlations. Simulation studies demonstrate devariation's superior performance over existing methods in various scenarios. We further validate devariation in multimodal neuroimaging data from the UK Biobank study, examining the associations between imaging-derived phenotypes (IDPs) from functional, structural, and diffusion magnetic resonance imaging (MRI).

stat.ME↗

arXiv2025

Joint Adaptive Penalty for Unbalanced Mediation Pathways

Mediation analysis has been widely used to investigate how a treatment influences an outcome through intermediate variables, known as mediators. Analyzing a mediation mechanism typically requires assessing multiple model parameters that characterize distinct pathwise effects. Classical methods that estimate these parameters individually can be inefficient, particularly when the underlying pathwise effects exhibit substantial imbalance. To address this challenge, this work proposes a new joint adaptive penalty that integrates information across entire mediation mechanisms, thereby enhancing both parameter estimation and pathway selection. We establish theoretical guarantees for the proposed method under an asymptotic framework and conduct extensive numerical studies to demonstrate its superior performance in scenarios with unbalanced mediation pathways.

stat.ME↗

arXiv2024

Quantile Mediation Analytics

Mediation analytics help examine if and how an intermediate variable mediates the influence of an exposure variable on an outcome of interest. Quantiles, rather than the mean, of an outcome are scientifically relevant to the comparison among specific subgroups in practical studies. Albeit some empirical studies available in the literature, there lacks a thorough theoretical investigation of quantile-based mediation analysis, which hinders practitioners from using such methods to answer important scientific questions. To address this significant technical gap, in this paper, we develop a quantile mediation analysis methodology to facilitate the identification, estimation, and testing of quantile mediation effects under a hypothesized directed acyclic graph. We establish two key estimands, quantile natural direct effect (qNDE) and quantile natural indirect effect (qNIE), in the counterfactual framework, both of which have closed-form expressions. To overcome the issue that the null hypothesis of no mediation effect is composite, we establish a powerful adaptive bootstrap method that is shown theoretically and numerically to achieve a proper type I error control. We illustrate the proposed quantile mediation analysis methodology through both extensive simulation experiments and a real-world dataset in that we investigate the mediation effect of lipidomic biomarkers for the influence of exposure to phthalates on early childhood obesity clinically diagnosed by 95\% percentile of body mass index.

stat.ME↗

arXiv2024

Efficient Analysis of Latent Spaces in Heterogeneous Networks

This work proposes a unified framework for efficient estimation under latent space modeling of heterogeneous networks. We consider a class of latent space models that decompose latent vectors into shared and network-specific components across networks. We develop a novel procedure that first identifies the shared latent vectors and further refines estimates through efficient score equations to achieve statistical efficiency. Oracle error rates for estimating the shared and heterogeneous latent vectors are established simultaneously. The analysis framework offers remarkable flexibility, accommodating various types of edge weights under general distributions.

stat.ME↗

arXiv2024

Selective inference for multiple pairs of clusters after K-means clustering

If the same data is used for both clustering and for testing a null hypothesis that is formulated in terms of the estimated clusters, then the traditional hypothesis testing framework often fails to control the Type I error. Gao et al. [2022] and Chen and Witten [2023] provide selective inference frameworks for testing if a pair of estimated clusters indeed stem from underlying differences, for the case where hierarchical clustering and K-means clustering, respectively, are used to define the clusters. In applications, however, it is often of interest to test for multiple pairs of clusters. In our work, we extend the pairwise test of Chen and Witten [2023] to a test for multiple pairs of clusters, where the cluster assignments are produced by K-means clustering. We further develop an analogous test for the setting where the variance is unknown, building on the work of Yun and Barber [2023] that extends Gao et al. [2022]'s pairwise test to the case of unknown variance. For both known and unknown variance settings, we present methods that address certain forms of data-dependence in the choice of pairs of clusters to test for. We show that our proposed tests control the Type I error, both theoretically and empirically, and provide a numerical study of their empirical powers under various settings.

stat.ME↗

arXiv2023

Extended Asymptotic Identifiability of Nonparametric Item Response Models

Nonparametric item response models provide a flexible framework in psychological and educational measurements. Douglas (2001) established asymptotic identifiability for a class of models with nonparametric response functions for long assessments. Nevertheless, the model class examined in Douglas (2001) excludes several popular parametric item response models. This limitation can hinder the applications in which nonparametric and parametric models are compared, such as evaluating model goodness-of-fit. To address this issue, We consider an extended nonparametric model class that encompasses most parametric models and establish asymptotic identifiability. The results bridge the parametric and nonparametric item response models and provide a solid theoretical foundation for the applications of nonparametric item response models for assessments with many items.

math.ST↗

arXiv2023

Rank-adaptive covariance testing with applications to genomics and neuroimaging

In biomedical studies, testing for differences in covariance offers scientific insights beyond mean differences, especially when differences are driven by complex joint behavior between features. However, when differences in joint behavior are weakly dispersed across many dimensions and arise from differences in low-rank structures within the data, as is often the case in genomics and neuroimaging, existing two-sample covariance testing methods may suffer from power loss. The Ky-Fan(k) norm, defined by the sum of the top Ky-Fan(k) singular values, is a simple and intuitive matrix norm able to capture signals caused by differences in low-rank structures between matrices, but its statistical properties in hypothesis testing have not been studied well. In this paper, we investigate the behavior of the Ky-Fan(k) norm in two-sample covariance testing. Ultimately, we propose a novel methodology, Rank-Adaptive Covariance Testing (RACT), which is able to leverage differences in low-rank structures found in the covariance matrices of two groups in order to maximize power. RACT uses permutation for statistical inference, ensuring an exact Type I error control. We validate RACT in simulation studies and evaluate its performance when testing for differences in gene expression networks between two types of lung cancer, as well as testing for covariance heterogeneity in diffusion tensor imaging (DTI) data taken on two different scanner types.

stat.ME↗

arXiv2023

Semiparametric Modeling and Analysis for Longitudinal Network Data

We introduce a semiparametric latent space model for analyzing longitudinal network data. The model consists of a static latent space component and a time-varying node-specific baseline component. We develop a semiparametric efficient score equation for the latent space parameter by adjusting for the baseline nuisance component. Estimation is accomplished through a one-step update estimator and an appropriately penalized maximum likelihood estimator. We derive oracle error bounds for the two estimators and address identifiability concerns from a quotient manifold perspective. Our approach is demonstrated using the New York Citi Bike Dataset.

math.ST↗

arXiv2020

A Note on the Likelihood Ratio Test in High-Dimensional Exploratory Factor Analysis

The likelihood ratio test is widely used in exploratory factor analysis to assess the model fit and determine the number of latent factors. Despite its popularity and clear statistical rationale, researchers have found that when the dimension of the response data is large compared to the sample size, the classical chi-square approximation of the likelihood ratio test statistic often fails. Theoretically, it has been an open problem when such a phenomenon happens as the dimension of data increases; practically, the effect of high dimensionality is less examined in exploratory factor analysis, and there lacks a clear statistical guideline on the validity of the conventional chi-square approximation. To address this problem, we investigate the failure of the chi-square approximation of the likelihood ratio test in high-dimensional exploratory factor analysis, and derive the necessary and sufficient condition to ensure the validity of the chi-square approximation. The results yield simple quantitative guidelines to check in practice and would also provide useful statistical insights into the practice of exploratory factor analysis.

math.ST↗

arXiv2020

On the Phase Transition of Wilk's Phenomenon

Wilk's theorem, which offers universal chi-squared approximations for likelihood ratio tests, is widely used in many scientific hypothesis testing problems. For modern datasets with increasing dimension, researchers have found that the conventional Wilk's phenomenon of the likelihood ratio test statistic often fails. Although new approximations have been proposed in high dimensional settings, there still lacks a clear statistical guideline regarding how to choose between the conventional and newly proposed approximations, especially for moderate-dimensional data. To address this issue, we develop the necessary and sufficient phase transition conditions for Wilk's phenomenon under popular tests on multivariate mean and covariance structures. Moreover, we provide an in-depth analysis of the accuracy of chi-squared approximations by deriving their asymptotic biases. These results may provide helpful insights into the use of chi-squared approximations in scientific practices.

math.ST↗

arXiv2020

Asymptotically Independent U-Statistics in High-Dimensional Testing

Many high-dimensional hypothesis tests aim to globally examine marginal or low-dimensional features of a high-dimensional joint distribution, such as testing of mean vectors, covariance matrices and regression coefficients. This paper constructs a family of U-statistics as unbiased estimators of the $\ell_p$-norms of those features. We show that under the null hypothesis, the U-statistics of different finite orders are asymptotically independent and normally distributed. Moreover, they are also asymptotically independent with the maximum-type test statistic, whose limiting distribution is an extreme value distribution. Based on the asymptotic independence property, we propose an adaptive testing procedure which combines $p$-values computed from the U-statistics of different orders. We further establish power analysis results and show that the proposed adaptive procedure maintains high power against various alternatives.

math.ST↗

arXiv2019

Likelihood Ratio Test in Multivariate Linear Regression: from Low to High Dimension

Multivariate linear regressions are widely used statistical tools in many applications to model the associations between multiple related responses and a set of predictors. To infer such associations, it is often of interest to test the structure of the regression coefficients matrix, and the likelihood ratio test (LRT) is one of the most popular approaches in practice. Despite its popularity, it is known that the classical $χ^2$ approximations for LRTs often fail in high-dimensional settings, where the dimensions of responses and predictors $(m,p)$ are allowed to grow with the sample size $n$. Though various corrected LRTs and other test statistics have been proposed in the literature, the fundamental question of when the classic LRT starts to fail is less studied, an answer to which would provide insights for practitioners, especially when analyzing data with $m/n$ and $p/n$ small but not negligible. Moreover, the power performance of the LRT in high-dimensional data analysis remains underexplored. To address these issues, the first part of this work gives the asymptotic boundary where the classical LRT fails and develops the corrected limiting distribution of the LRT for a general asymptotic regime. The second part of this work further studies the test power of the LRT in the high-dimensional setting. The result not only advances the current understanding of asymptotic behavior of the LRT under alternative hypothesis, but also motivates the development of a power-enhanced LRT. The third part of this work considers the setting with $p>n$, where the LRT is not well-defined. We propose a two-step testing procedure by first performing dimension reduction and then applying the proposed LRT. Theoretical properties are developed to ensure the validity of the proposed method. Numerical studies are also presented to demonstrate its good performance.

math.ST↗

arXiv2018

Estimating Tail Probabilities of the Ratio of the Largest Eigenvalue to the Trace of a Wishart Matrix

This paper develops an efficient Monte Carlo method to estimate the tail probabilities of the ratio of the largest eigenvalue to the trace of the Wishart matrix, which plays an important role in multivariate data analysis. The estimator is constructed based on a change-of-measure technique and it is proved to be asymptotically efficient for both the real and complex Wishart matrices. Simulation studies further show the outperformance of the proposed method over existing approaches based on asymptotic approximations, especially when estimating probabilities of rare events.

stat.ME↗