Incoming DPhil in Primary Care Health Sciences, University of Oxford (from Oct 2026). MPhil in Population Health Sciences, University of Cambridge (2025–2026). Genetic epidemiology and causal inference, and turning hard statistical methods into things people can actually read.

I work mostly in R, care about reproducibility, and like explaining a method well enough that a non-specialist walks away understanding it.

LinkedIn · yaonyyao@gmail.com · ORCID

• Genetic epidemiology: rare-variant association analysis, penalised (Firth) logistic regression for imbalanced case–control data
• Causal inference: Mendelian randomisation, instrumental variables, sensitivity analysis for pleiotropy
• Evaluation methodology: subgroup performance, calibration, and whether a study was ever powered to find what it claims to have looked for
• Reproducible research: Quarto, seeded simulation, hand-checked estimators validated against established packages

Mendelian Randomisation from Scratch » A seven-chapter R walkthrough that codes the standard MR toolkit (2SLS, IVW, MR-Egger, weighted median) from first principles on synthetic data, then validates each estimator against ivreg / MendelianRandomization. https://lexiyao.github.io/mendelian-randomisation-from-scratch/

Odds-Ratio Stats Converter » A dependency-free, single-file web tool for the quantities you meet reading genetic-epi and GWAS results (OR, log-OR, SE, confidence interval and two-sided p), recomputing everything live from the canonical log-OR and its standard error, with a paste-a-reported-string parser. Math core unit-tested against known values. https://lexiyao.github.io/or-stats-converter/

Evidence Briefs » Two-page, plain-language evidence summaries that turn a body of causal-inference and clinical-trial evidence into something a non-specialist can act on: bottom line first, then the certainty and the limits. Worked example: does lowering LDL cholesterol reduce coronary heart disease? https://lexiyao.github.io/evidence-brief-ldl-chd/

Current: MPhil dissertation on the genetic architecture of benign breast disease, DCIS and LCIS and their links to breast-cancer susceptibility genes (private while under examination).

R · Quarto · tidyverse / ggplot2 · Mendelian randomisation · GWAS · survival & logistic regression · prediction-model evaluation · reproducible pipelines