課程名稱 
統計思考 Statistical Thinking 
開課學期 
1012 
授課對象 
公共衛生學院 流預所生物醫學統計組 
授課教師 
杜裕康 
課號 
EPM5001 
課程識別碼 
849 U0300 
班次 

學分 
2 
全/半年 
半年 
必/選修 
選修 
上課時間 
星期一5,6(12:20~14:10) 
上課地點 
公衛215 
備註 
限研究生和大三，大四學生選修。 總人數上限：15人 
Ceiba 課程網頁 
http://ceiba.ntu.edu.tw/1012sthinking2013 
課程簡介影片 

核心能力關聯 
本課程尚未建立核心能力關連 
課程大綱

為確保您我的權利,請尊重智慧財產權及不得非法影印

課程概述 
It is essential for graduate students to develop critical thinking in order to identify caveat in current research, develop new approaches to existing research problems and explore new horizon in their research fields. Students would find it very helpful, if real examples in clinical and epidemiological research are used to demonstrate how to achieve these goals, and this is the aim of this course. The syllabus of this course will loosely follow the order of chapters in my new book of the same title published by Chapman & Hall this summer with supplementary materials. Two tools to promote understanding of statistical modelling will be first introduced: vector geometry for linear models and the directed acyclic graphs for causal thinking. Then, background knowledge will be provided for students to identify the potential caveats and to use their previous knowledge to choose the most appropriate approaches or to develop new approaches. Potential scenarios and examples used in this course include:
• Testing the relation between baseline and changes
• The use of ratio variables in regression analysis
• Statistical methods for testing differences in changes in randomised controlled trials
• Lord’s paradox and the adjustment of baseline values in observational studies
• The problem of collinearity in linear models
• Testing statistical and biological interaction
• Confounding, causality and Simpson’s paradox
• Reversal paradox and the adjustment of intermediate variables on a causal pathway
• Testing direct, indirect and total effects
Students will be asked to give presentations to discuss the reading materials and lead discussion. Students’ performance will be assessed by their participation in the classroom discussion, their presentations and the final essays on one of the topics discussed in the course.

課程目標 
By the end of this course, students should be able to:
• Describe the rationales of directed acyclic graphs in causal inference
• Describe how to use vector geometry to represent linear models
• Describe the problems with mathematical coupling in testing associations between variables
• Explain the problem of regression to the mean with assessing the impact of baseline measurements on changes from baseline
• Explain the potential problems caused by imbalance in baseline covariates in assessing group differences in changes from baseline
• Describe the strength and limitations in methods for adjusting baseline imbalance
• Describe the conceptual relations between biological and statistical interaction
• Explain the causes and potential solutions to the problem of reversal paradox
• Explain the different approaches to separate direct and indirect effects and their underlying assumptions

課程要求 
Students are required to do a midterm and final presentation on their projects and submit final group project at the end of the course. 
預期每週課後學習時數 

Office Hours 
備註： 10% class participation + 30% presentation + 60% final report 
指定閱讀 
• Journal articles
• YuKang Tu, Mark Gilthorpe: “Statistical Thinking in Epidemiology”. Chapman & Hall, 2012.

參考書目 
• Judea Pearl. “Causality” 2nd edition. Cambridge University Press, 2010.
• Bjorn Andersen. Methodological errors in medical research. London: Blackwell, 2000.
• Steven Sloman. “Causal Models”. Oxford University Press, 2005.

評量方式 (僅供參考) 

週次 
日期 
單元主題 
第1週 
2/18 
What is “statistical thinking”? 
第2週 
2/25 
Vector geometry for linear models (1) 
第3週 
3/04 
Vector geometry for linear models (2) 
第4週 
3/11 
Causation and directed acyclic graphs (1) 
第5週 
3/18 
國際學術交流 
第6週 
3/25 
Causation and Directed Acyclic Graph (2) 
第7週 
4/01 
Lord's paradox 
第8週 
4/08 
Seminar: why is randomized controlled trial the “gold standard”? 
第9週 
4/15 
Seminar: Mendelian Randomization 
第10週 
4/22 
Special lecture: causal inference from an econometrics perspective (1) (鄭凱文老師) 
第11週 
4/29 
Special lecture: causal inference from an econometrics perspective (2) (鄭凱文老師) 
第12週 
5/06 
Seminar: Propensity scores 
第13週 
5/13 
Collinearity & reversal paradox 
第14週 
5/20 
The law of initial value 
第15週 
5/27 
Interaction 
第16週 
6/03 
Final student presentation (1) 
第17週 
6/10 
Final student presentation (2) 
