Research

Robust Sensitivity Analysis for Quantiles of Individual Treatment Effects in Paired Observational Studies

advised by Xinran Li, University of Illinois at Urbana-Champaign (Ongoing)

  • Generalize sensitivity analysis to deal with quantiles of individual treatment effects and quantiles of the hidden biases from all matched sets.
  • Develop an adaptive test in observational study, together with simulations showing the robustness compare to the usual tests.
  • Developing R package.

Sensitivity Analysis for Quantiles of Hidden Biases in Matched Observational Studies

advised by Xinran Li, University of Illinois at Urbana-Champaign (July 2020 – July 2021)

  • Generalize Rosenbaum’s conventional framework to conduct sensitivity analysis on quantiles of hidden biases from all matched sets, which are more robust than the maximum.
  • Demonstrate that the proposed sensitivity analysis for all quantiles of hidden biases is simultaneously valid and is thus a free lunch added to the conventional sensitivity analysis.
  • An R package of our method is available.
  • Paper has submitted to Journal of the American Statistical Association.

Simple and Fast Algorithms for Rank Estimation of Right-censored Length-biased Data

advised by Gary Chan, University of Washington (June 2018 – Sept. 2018)

  • Implemented a fast algorithm in R to solve an over-identified set of non-smooth and non-monotone log-rank estimating equations based on the left-truncated weight-censored data and backward recurrence time.

Recommendation System on Traditional Chinese Medicine Big Data 

advised by Xiaohua Zhou, Peking University (June 2017 – June 2018)

  • Implemented the recommendation systems through neighborhood-based and model-based collaborative filtering, and deep learning based techniques such as item embedding, feedforward networks, auto-encoders.
  • Constructed hybrid recommendation system based on DeepFM and neighborhood-based model and tested on real data.