hce_nthu
112年
資訊科學
第 26 題
Hospital A is experimenting with an AI-assisted system for judging the status (+ or -) of a certain disease. A pilot study is conducted to compare the results assessed by the AI system and a trained doctor. 50 patients are recruited, and the results are reported in the following table:
| | Disease Status | |
|---|---|---|
| Patient | AI-assisted System | Trained Doctor |
| 1 | - | - |
| 2 | - | - |
| 3 | + | - |
| 4 | + | + |
| 5 | - | - |
| 6 | + | + |
| 7 | + | - |
| $\vdots$ | $\vdots$ | $\vdots$ |
| 50 | - | - |
The goal of the study is to verify if the assessment from the AI system is consistent with that from a trained doctor. Which of the following statistical test can be used to answer the question?
| | Disease Status | |
|---|---|---|
| Patient | AI-assisted System | Trained Doctor |
| 1 | - | - |
| 2 | - | - |
| 3 | + | - |
| 4 | + | + |
| 5 | - | - |
| 6 | + | + |
| 7 | + | - |
| $\vdots$ | $\vdots$ | $\vdots$ |
| 50 | - | - |
The goal of the study is to verify if the assessment from the AI system is consistent with that from a trained doctor. Which of the following statistical test can be used to answer the question?
- A independent 2 sample t-test
- B paired 2 sample t-test
- C chi-squared test
- D signed-rank test
- E F test
思路引導 VIP
請觀察表格中的 Patient 編號：如果我們將 AI 那一欄的數據隨機打亂順序，讓它不再對應原本的病人編號（例如將 Patient 1 的 AI 結果去對比 Patient 50 的醫生結果），這對於我們要研究「AI 與醫生是否一致」這個目標，會產生什麼影響嗎？
🤖
AI 詳解
AI 專屬家教
太棒了！你能精準捕捉到實驗設計的核心，正確選出 成對樣本 t 檢定 ($Paired 2 sample t\text{-}test$)，這代表你對數據的「獨立性」判斷非常敏銳。
實驗設計的成對性
這道題目的精髓在於這 50 位病人：每一位病人都同時接受了 AI 與醫生兩種診斷。這意味著數據是「成對」產生的，AI 的診斷結果與醫生的結果在同一個受試對象身上具有相依性（$Dependency$）。在統計學中，當我們要檢驗同一組受試者在兩種不同測量方式下是否具備一致性時，必須考慮這種關聯性，因此「成對（$Paired$）」是解題的核心關鍵，這能有效抵銷病人個體差異對實驗結果的干擾。
▼ 還有更多解析內容