觀察性研究的因果推論：給臨床醫師的實戰指南

Causal Inference from Observational Studies: A Practical Guide for Clinicians

Live Demo: htlin222.github.io/causal-tut

課程內容

章節	主題	重點
1-2	問題與設計	為什麼需要因果推論？Target Trial 設計
3-4	方法概念	PS → IPW → 雙重穩健 → TMLE
5-6	準備工作	選方法、備資料
7	R 實作	程式碼示範
8-9	驗證報告	E-value、論文寫作

Features

• reveal-auto-agenda - Generates agenda slides automatically
• code-fullscreen - Fullscreen button for code blocks
• codeFocus - Progressive code line highlighting with explanations
• simplemenu - Navigation menu bar for sections

R Scripts

Runnable analysis scripts in Rscript/chapter_09_r-workflow/:

Script	Description	Outputs
01_load_data.R	Load and inspect datasets	-
02_tmle_continuous.R	TMLE for continuous outcome	HTML tables, forest plot
03_tmle_binary.R	TMLE for binary outcome (RD/RR)	HTML tables, forest plot
04_survival_tmle.R	Survival TMLE	HTML tables, KM curve
05_survival_ipw_cox.R	IPW-weighted Cox regression	HTML tables, forest plot, KM curve
06_balance_diagnostics.R	Covariate balance diagnostics	Love plot, PS overlap, weight distribution

Run from repo root:

Rscript Rscript/chapter_09_r-workflow/02_tmle_continuous.R

Outputs saved to output/chapter_09/ (gitignored).

Quick Start

# Build both formats (slides.html + index.html)
make

# Live preview with auto-reload
make preview

# Clean output files
make clean

Note: index.qmd includes chapters/*.qmd via Quarto include syntax.

Output

File	Description	URL
index.html	HTML page with TOC	htlin222.github.io/causal-tut/
slides.html	Revealjs presentation	htlin222.github.io/causal-tut/slides.html

Requirements

• Quarto ≥ 1.3
• R packages: tmle, survtmle, SuperLearner, WeightIt, cobalt, gtsummary, gt, ggplot2, ggsurvfit, survival

License

MIT