R語言臨床預測模型實戰

本書從R語言臨床預測模型基本原理講起，逐步深入到進階實戰，並配合實戰案例，重點介紹臨床預測模型的構建、評價、驗證，讓讀者可以快速領悟 3~5分預測模型 SCI（science citation index，科學引文索引）的套路與精髓，為後續沖擊10分以上SCI打基礎。 本書分為13章，主要內容有線性回歸、Logistic回歸、Cox回歸、競爭風險模型等；自變量篩選方法有傳統方法、逐步回歸法、Lasso法、隨機森林法、**子集法、主成分分析法等；模型可視化涉及多種形式的列線圖、Calibration校準曲線、ROC、DCA 曲線等圖形繪制，不僅涉及單模型的可視化，還涉及單模型多時點、多模型同時點的可視化；模型評價指標涉及C指數、AUC、NRI、IDI 等；模型驗證主要涉及簡單交叉驗證、K折交叉驗證、留一法交叉驗證及Bootstrap法。 本書內容通俗易懂，案例豐富，實用性強，適合R語言臨床預測模型的入門讀者和進階讀者，尤其是臨床、護理、公共衛生、藥學等專業的碩士研究生和博士研究生或從事相關研究的科研人員閱讀。另外，本書還可以作為相關培訓機構的教材。

目錄

第1章 臨床預測模型概述 ············································1

1.1 如何構建預測模型 ·································································2

1.1.1 先單後多 ······················································································2

1.1.2 逐步回歸 ······················································································2

1.1.3 Lasso ···························································································2

1.1.4 隨機森林 ······················································································3

1.1.5 最優子集 ······················································································3

1.2 如何評價預測模型 ·································································3

1.2.1 擬合優度檢驗 ················································································3

1.2.2 ROC ····························································································4

1.2.3 Calibration ·····················································································4

1.2.4 DCA曲線 ·····················································································4

1.3 如何驗證預測模型 ·································································4

1.3.1 交叉驗證 ······················································································5

1.3.2 Bootstrap法 ···················································································5

1.4 小結 ···················································································5

第2章 線性回歸 ························································6

2.1 線性回歸概述 ·······································································7

2.1.1 相關的概念 ···················································································7

2.1.2 回歸的概念 ···················································································9

2.2 線性回歸實戰 ······································································11

2.2.1 簡單線性回歸 ···············································································11

IV

2.2.2 多重線性回歸 ···············································································17

2.3 小結 ··················································································25

第3章 Logistic回歸 ··················································27

3.1 概述 ··················································································28

3.1.1 二分類Logistic回歸的基本原理························································28

3.1.2 多分類Logistic回歸的基本原理························································30

3.1.3 有序Logistic回歸的基本原理···························································30

3.1.4 1∶m匹配條件Logistic回歸的基本原理···········································31

3.2 Logistic回歸分析實戰 ···························································31

3.2.1 二分類Logistic回歸 ·······································································31

3.2.2 多分類Logistic回歸 ·······································································41

3.2.3 有序Logistic ················································································49

3.2.4 1∶m匹配條件Logistic ·································································53

3.3 小結 ··················································································56

第4章生存資料分析 ················································59

4.1 概述 ··················································································60

4.1.1 Kaplan-Meier生存曲線的基本原理····················································60

4.1.2 Cox比例風險模型的基本原理 ··························································61

4.2 生存資料分析實戰 ································································62

4.2.1 Kaplan-Meier生存曲線 ···································································62

4.2.2 Cox比例風險模型 ·········································································72

4.3 小結 ··················································································81

第5章競爭風險模型 ················································83

5.1 概述 ··················································································84

5.2 競爭風險模型實戰 ································································85

5.2.1 Fine-Gray檢驗（單因素分析） ·························································85

5.2.2 crr()函數法（多因素分析） ·····························································95

V

目錄

5.3 小結 ··················································································97

第6章 自變量篩選 ···················································99

6.1 傳統方法 ··········································································100

6.1.1 基於連續性資料 ··········································································100

6.1.2 基於二分類資料 ··········································································105

6.1.3 基於生存資料 ·············································································112

6.2 逐步法 ·············································································116

6.2.1 基於連續性資料 ··········································································117

6.2.2 基於二分類資料 ··········································································121

6.2.3 基於多分類資料 ··········································································124

6.2.4 基於有序資料 ·············································································126

6.2.5 基於1∶m匹配資料 ···································································128

6.2.6 基於生存資料 ·············································································130

6.2.7 基於競爭風險資料 ·······································································131

6.3 Lasso法 ···········································································134

6.3.1 基於連續性資料 ··········································································135

6.3.2 基於二分類資料 ··········································································144

6.3.3 基於多分類資料 ··········································································151

6.3.4 基於生存資料 ·············································································160

6.4 隨機森林法 ·······································································167

6.4.1 基於連續性資料 ··········································································168

6.4.2 基於二分類資料 ··········································································172

6.4.3 基於多分類資料 ··········································································177

6.4.4 基於生存資料 ·············································································180

6.4.5 基於競爭風險資料 ·······································································187

6.5 最優子集法 ·······································································191

6.5.1 基於連續性資料 ··········································································192

6.5.2 基於二分類資料 ··········································································193

6.5.3 基於多分類資料 ··········································································193

6.5.4 基於有序資料 ·············································································196

6.5.5 基於生存資料 ·············································································196

VI

6.5.6 基於競爭風險資料 ·······································································198

6.6 主成分分析法 ····································································200

6.7 小結 ················································································204

第7章列線圖·························································205

7.1 列線圖簡介 ·······································································206

7.2 基於連續資料 ····································································206

7.2.1 普通靜態列線圖 ··········································································206

7.2.2 網頁動態列線圖 ··········································································210

7.2.3 花樣列線圖 ················································································212

7.2.4 靜態列線圖的變種 ·······································································215

7.3 基於二分類資料 ·································································216

7.3.1 普通靜態列線圖 ··········································································216

7.3.2 網頁動態列線圖 ··········································································222

7.3.3 花樣列線圖 ················································································223

7.3.4 靜態列線圖的變種 ·······································································227

7.4 基於有序資料 ····································································228

7.4.1 普通靜態列線圖 ··········································································228

7.4.2 花樣列線圖 ················································································233

7.4.3 靜態列線圖的變種 ·······································································236

7.5 基於生存資料 ····································································237

7.5.1 普通靜態列線圖 ··········································································237

7.5.2 網頁動態列線圖 ··········································································246

7.5.3 花樣列線圖 ················································································248

7.5.4 靜態列線圖的變種 ·······································································253

7.5.5 生存概率列線圖 ··········································································255

7.6 基於競爭風險模型 ······························································256

7.6.1 普通靜態列線圖 ··········································································256

7.6.2 靜態列線圖的變種 ·······································································264

7.6.3 生存概率列線圖 ··········································································265

7.6.4 花樣列線圖 ················································································266

7.7 小結 ················································································270

第8章 Calibration校準曲線 ·································271

8.1 Calibration校準曲線簡介 ·····················································272

8.2 基於二分類資料 ·································································272

8.2.1 單一模型校準曲線 ·······································································272

8.2.2 多模型校準曲線 ··········································································281

8.3 基於生存資料 ····································································284

8.3.1 單模型單時點校準曲線 ·································································285

8.3.2 單模型多時點校準曲線 ·································································292

8.3.3 多模型同時點校準曲線 ·································································294

8.4 基於競爭風險模型 ······························································299

8.4.1 單模型單時點校準曲線 ·································································299

8.4.2 多模型同時點校準曲線 ·································································301

8.5 小結 ················································································304

第9章 C指數計算 ···················································305

9.1 C指數簡介 ·······································································306

9.2 基於二分類資料 ·································································306

9.3 基於生存資料 ····································································310

9.4 基於競爭風險模型 ······························································315

9.5 小結 ················································································318

第10章 ROC曲線 ···················································319

10.1 ROC曲線簡介 ·································································320

10.2 基於二分類資料 ·······························································320

10.2.1 單一模型ROC曲線 ····································································320

10.2.2 多模型ROC曲線 ·······································································328

10.3 基於生存資料 ··································································332

10.3.1 單模型單時點ROC曲線 ······························································332

10.3.2 單模型多時點ROC曲線 ······························································336

10.3.3 多模型同時點ROC曲線 ······························································339

VIII

10.4 基於競爭風險模型····························································343

10.4.1 單模型單時點ROC曲線 ······························································343

10.4.2 單模型多時點ROC曲線 ······························································345

10.4.3 多模型同時點ROC曲線 ······························································348

10.5 小結 ··············································································349

第11章 DCA曲線 ···················································350

11.1 DCA曲線簡介 ·································································351

11.2 基於二分類資料·······························································351

11.2.1 單一模型DCA曲線····································································351

11.2.2 多模型DCA曲線·······································································356

11.3 基於生存資料··································································359

11.3.1　單模型單時點DCA曲線 ······························································359

11.3.2　單模型多時點DCA曲線 ······························································363

11.3.3　多模型同時點DCA曲線 ······························································365

11.4 小結 ··············································································368

第12章 NRI、IDI計算 ·············································369

12.1 NRI、IDI簡介 ·································································370

12.2 基於二分類資料 ·······························································370

12.3 基於生存資料 ··································································378

12.4 小結 ··············································································384

第13章交叉驗證及Bootstrap ·····································385

13.1 概述 ··············································································386

13.2 簡單交叉驗證 ··································································389

13.3 K折交叉驗證 ··································································391

13.4 留一法交叉驗證 ·······························································393

13.5 Bootstrap ·········································································395

13.6 小結 ··············································································397