Research

Almost all statistical outputs, including estimation, testing and prediction, are “wrong”.

Quantifying how wrong they are is the spirit of statistical inference. Variance estimation arguably plays a central role in inference. This problem connects a large portion of my research projects. In recent years, I am generally interested in statistical inference for dependent data and incomplete data. I am also interested in elegant statistical theories and methodologies in a wide variety of areas, e.g., non-parametric methods, robust analysis and fiducial inference.

Inference for Dependent Data

• • • Estimation of asymptotic variance

    • Change-point and trend detection

    • Recursive estimation

Inference for Incomplete Data

• • • Multiple imputation

    • Estimation of fraction of missing information

    • Multi-phase inference

General Theory & Methodology

• • • Non-parametric methods

    • Fiducial inference

    • Hypothesis testing

    • Robust methods

    • Subsampling methods