Portfolio item number 1
Short description of portfolio item number 1
Posts by Collection
portfolio
Portfolio item number 2
Short description of portfolio item number 2
publications
Model-free, monotone invariant and computationally efficient feature screening with data-adaptive threshold
Published in Journal of Statistical Planning and Inference, 2024
We propose SIT-BY, a model-free, efficient feature screening method for ultrahigh-dimensional data. By using sliced independence estimates and an adaptive FDR-based threshold, SIT-BY controls false discoveries while retaining all relevant features, as demonstrated in simulations and real data.
Structure-adaptive canonical correlation analysis for microbiome multi-omics data
Published in Frontiers in Genetics, 2024
We introduce a new sCCA framework for integrating microbiome and other high-dimensional omics data, explicitly accounting for the compositional nature of microbiome data. Our method incorporates prior structural information among taxa via adaptive penalization, improving accuracy when structure is informative and remaining robust to misspecification. Extensive simulations and real data analyses show that this approach outperforms existing methods.
Joint mirror procedure: Controlling false discovery rate for identifying simultaneous signals
Published in Biometrics, 2024
We present the joint mirror (JM) procedure for detecting features with joint statistical significance while controlling the false discovery rate (FDR) in finite samples. JM iteratively shrinks the rejection region based on updated information until the estimated false discovery proportion meets the target FDR. We further introduce the composite FDR (cFDR), which weights false discoveries by their number of null components, and prove JM’s control of cFDR using a leave-one-out technique. Our efficient algorithm accommodates partial ordering and, through simulations and real applications, demonstrates robust FDR control and improved power.
Multiplex depth for network-valued data and applications
Published in Journal of Computational and Graphical Statistics, 2025
We introduce multiplex depth, a novel extension of Tukey’s depth for analyzing network-valued data. It enables data ordering, center estimation, and outlier detection. We establish its theoretical properties, develop efficient algorithms, and validate our method through simulations and brain network data, demonstrating improved performance on key analysis tasks.
A general framework for modeling Gaussian process with qualitative and quantitative factors
Published in Technometrics, 2026
We develop a general latent-variable Gaussian process framework for computer experiments involving qualitative and quantitative factors. The framework accommodates standard kernels and ordinal structures, and supports model selection and averaging using BIC and leave-one-out cross-validation.
Powerful spatial multiple testing via borrowing neighboring information
Published in Statistica Sinica, 2026
We propose 2d-SMT, a multiple testing method that leverages spatial information via an auxiliary statistic to improve FDR control and power for spatial signals. 2d-SMT integrates with weighted BH procedures, offers fast optimization, and shows superior FDR-power performance in theory and simulations.
teaching
Probability Theory
Graduate course, Renmin University of China, Institute of Statistics and Big Data, 2020
Statistical Learning II
Graduate course, Renmin University of China, Institute of Statistics and Big Data, 2021
Algorithm Design and Analysis
Undergraduate course, Renmin University of China, Institute of Statistics and Big Data, 2022
Algorithm Design and Analysis
Graduate course, Renmin University of China, Institute of Statistics and Big Data, 2022
Advanced Statistical Computing
Graduate course, Renmin University of China, Institute of Statistics and Big Data, 2023
Statistical Learning
Undergraduate course, Renmin University of China, Institute of Statistics and Big Data, 2024