Introduction:

Uncertainty exists in many real-life problems, ranging from stock returns to sport results to medication effects to election outcomes. Statistics offers methods to handle uncertainty with a higher precision. Improving a decision with 50-50 certainty to 60-40 certainty makes a huge difference in many practically problems. This course introduces ways to (i) define, (ii) model and (iii) forecast uncertainty through real-life examples and counterintuitive phenomena. Topics include (i) birthday paradox, Simpson’s paradox; (ii) linear regression model, auto-regressive regression model, logistic regression model, non-parametric regression model; and (iii) historical simulation, and k-mean clustering.

不確定性存在於許多現實生活中的問題，例子涵蓋股票回報、運動結果、藥物效果、選舉結果等。統計科學提供了具更高準定性的方法，以處理不確定性的問題。在許多實際問題中，將50-50的不確定性提高至60-40，可令數據分析變得更精確。本課程以實際示例和違反直覺的現象引導，來 (i) 定義、(ii) 模型和 (iii) 預測不確定性。主題包括 (i)生日悖論，辛普森悖論；(ii) 線性迴歸模型，自迴歸模型，邏輯迥歸模型，非參數迴歸模型；(iii) 歷史模擬法，k平均演算法。