01 Home02 Research03 Projects04 Notes05 About06 Lab
Home/Research

Questions first.
Methods in their service.

My work connects statistical reasoning to practical questions in health, computation and financial decisions. Four directions, each with the setting it came from and where its evidence stops.

01 · Causal inference

Health data, mediation and time.

At HKU, in Professor Dora Y. Zhang's research group, I work on statistical questions involving health-related exposures and disease outcomes. My ongoing longitudinal work asks how to distinguish between-person differences from within-person change, and what assumptions a mediation claim needs.

Setting
Dora Y. Zhang's group, HKU
Data
China Family Panel Studies
Status
In progress

An association is not presented as a causal effect without an identified estimand and a sensitivity analysis; no estimates are published here yet.

Project: multi-wave mediation in CFPS →
CFPS study design over four survey waves, conceptual.CFPS · study design, conceptual
02 · Statistical machine learning

Evaluating generative models beyond appearance.

In Li Ma's research group at Duke, I studied generative models for microbiome data and evaluated generated samples using bioinformatics analyses and large-scale computing.

Setting
Li Ma's group, Duke University
Period
May – Dec 2024
Role
Collaborative research

Research experience; the tile beside it is a schematic, not results.

Earlier research on the Projects page →
Schematic: observed and generated microbiome compositions; not data.Schematic · not data
03 · Optimization

Optimization under uncertainty.

In Ying Cui's research group at UC Berkeley, I worked on a semismooth Newton approach to risk-aware, nonconvex optimization and compared its computational behaviour with other solvers.

Setting
Ying Cui's group, UC Berkeley
Role
Collaborative research

Research experience; the tile beside it is a schematic of iterates on level sets, not results.

Earlier research on the Projects page →
Schematic: Newton-type iterates converging on level sets; not data.Schematic · not data
04 · Quantitative research

From prediction to a defensible decision.

At Fudan, I studied deep learning price forecasts with an A-share backtest and monthly portfolio rebalancing. My current independent projects focus on point-in-time data, honest out-of-sample evaluation and trading frictions.

Earlier
Weiping Zhang's group, Fudan, 2023
Now
Helix and the Kronos evaluation
Status
Ongoing

The screenshot is a saved in-sample snapshot whose book fails Helix’s own validation gate; it is not a performance claim.

Case studies: Helix and Kronos →
The Helix research terminal, a saved snapshot of an in-sample backtest.Helix · research terminal, snapshot 27 Jul 2026

Learning by explaining.

Background and training →

Quantitative Strategies and Algorithmic Trading

Teaching duties · The University of Hong Kong

STAT8020

Spatial Data Analysis

Teaching duties · The University of Hong Kong

STAT6016