Efnisyfirlit

• Cover
• Half Title
• Series Page
• Title Page
• Copyright Page
• Dedication
• Contents
• Foreword
• Preface
• Terminology
• Software
• 1 Why should I care about statistical prediction models?
• 1.1 The many uses of prediction models in medicine
• 1.2 The unique messages of this book
• 1.3 Prognostic factor modeling philosophy
• 1.4 The rest of this book
• 2 I am going to make a prediction model. What do I need to know?
• 2.1 Prediction model framework
• 2.1.1 Target population
• 2.1.2 The time origin
• 2.1.3 The event of interest
• 2.1.4 The prediction time horizon and follow-up
• 2.1.5 Landmarking
• 2.1.6 Risks and risk predictions
• 2.1.7 Classification of risk
• 2.1.8 Predictor variables
• 2.1.9 Checklist
• 2.2 Prediction performance
• 2.2.1 Proper scoring rules
• 2.2.2 Calibration
• 2.2.3 Discrimination
• 2.2.4 Explained variation
• 2.2.5 Variability and uncertainty
• 2.2.6 The interpretation is relative
• 2.2.7 Utility
• 2.2.8 Average versus subgroups
• 2.3 Study design
• 2.3.1 Study design and sources of information
• 2.3.2 Cohort
• 2.3.3 Multi-center study
• 2.3.4 Randomized clinical trial
• 2.3.5 Case-control
• 2.3.6 Given treatment and treatment options
• 2.3.7 Sample size calculation
• 2.4 Data
• 2.4.1 Purpose dataset
• 2.4.2 Data dictionary
• 2.4.3 Measurement error
• 2.4.4 Missing values
• 2.4.5 Censored data
• 2.4.6 Competing risks
• 2.5 Modeling
• 2.5.1 Risk prediction model
• 2.5.2 Risk classifier
• 2.5.3 How is prediction modeling different from statistical inference?
• 2.5.4 Regression model
• 2.5.5 Linear predictor
• 2.5.6 Expert selects the candidate predictors
• 2.5.7 How to select variables for inclusion in the final model
• 2.5.8 All possible interactions
• 2.5.9 Checklist
• 2.5.10 Machine learning
• 2.6 Validation
• 2.6.1 The conventional model
• 2.6.2 Internal and external validation
• 2.6.3 Conditional versus expected performance
• 2.6.4 Cross-validation
• 2.6.5 Data splitting
• 2.6.6 Bootstrap
• 2.6.7 Model checking and goodness of fit
• 2.6.8 Reproducibility
• 2.7 Pitfalls
• 2.7.1 Age as time scale
• 2.7.2 Odds ratios and hazard ratios are not predictions of risks
• 2.7.3 Do not blame the metric
• 2.7.4 Censored data versus competing risks
• 2.7.5 Disease-specific survival
• 2.7.6 Overfitting
• 2.7.7 Data-dependent decisions
• 2.7.8 Balancing data
• 2.7.9 Independent predictor
• 2.7.10 Automated variable selection
• 3 How should I prepare for modeling?
• 3.1 Definition of subjects
• 3.2 Choice of time scale
• 3.3 Pre-selection of predictor variables
• 3.4 Preparation of predictor variables
• 3.4.1 Categorical variables
• 3.4.2 Continuous variables
• 3.4.3 Derived predictor variables
• 3.4.4 Repeated measurements
• 3.4.5 Measurement error
• 3.4.6 Missing values
• 3.5 Preparation of event time outcome
• 3.5.1 Illustration without competing risks
• 3.5.2 Illustration with competing risks
• 3.5.3 Artificial censoring at the prediction time horizon
• 4 I am ready to build a prediction model
• 4.1 Specifying the model type
• 4.1.1 Uncensored binary outcome
• 4.1.2 Right-censored time-to-event outcome (no competing risks)
• 4.1.3 Right-censored time-to-event outcome with competing risks
• 4.2 Benchmark model
• 4.2.1 Uncensored binary outcome
• 4.2.2 Right-censored time-to-event outcome (without competing risks)
• 4.2.3 Right-censored time-to-event with competing risks
• 4.3 Including predictor variables
• 4.3.1 Categorical predictor variables
• 4.3.2 Continuous predictor variables
• 4.3.3 Interaction effects
• 4.4 Modeling strategy
• 4.4.1 Variable selection
• 4.4.2 Conventional model strategy
• 4.4.3 Whether to use a standard regression model or something else
• 4.5 Advanced topics
• 4.5.1 How to prevent overfitting the data
• 4.5.2 How to deal with missing values
• 4.5.3 How to deal with non-converging models
• 4.6 What you should put in your manuscript
• 4.6.1 Baseline tables
• 4.6.2 Follow-up tables
• 4.6.3 Regression tables
• 4.6.4 Risk plots
• 4.6.5 Nomograms
• 4.7 Deployment
• 4.7.1 Risk charts
• 4.7.2 Internet calculator
• 4.7.3 Cost-benefit analysis (waiting lists)
• 5 Does my model predict accurately?
• 5.1 Model assessment roadmap
• 5.1.1 Visualization of the predictions
• 5.1.2 Calculation of model performance
• 5.1.3 Visualization of model performance
• 5.2 Uncensored binary outcome
• 5.2.1 Distribution of the predicted risks
• 5.2.2 Brier score
• 5.2.3 AUC
• 5.2.4 Calibration curves
• 5.3 Right-censored time-to-event outcome (without competing risks)
• 5.3.1 Distribution of the predicted risks
• 5.3.2 Brier score with censored data
• 5.3.3 Time-dependent AUC for censored data
• 5.3.4 Calibration curve for censored data
• 5.4 Competing risks
• 5.4.1 Distribution of the predicted risks
• 5.4.2 Brier score with competing risks
• 5.4.3 Time-dependent AUC for competing risks
• 5.4.4 Calibration curve for competing risks
• 5.5 The Index of Prediction Accuracy (IPA)
• 5.6 Choice of prediction time horizon
• 5.7 Time-dependent prediction performance
• 6 How do I decide between rival models?
• 6.1 Model comparison roadmap
• 6.2 Analysis of rival prediction models
• 6.2.1 Uncensored binary outcome
• 6.2.2 Right-censored time-to-event outcome (without competing risks)
• 6.2.3 Competing risks
• 6.3 Clinically relevant change of prediction
• 6.4 Does a new marker improve prediction?
• 6.4.1 Many new predictors
• 6.4.2 Updating a subject’s prediction
• 7 What would make me an expert?
• 7.1 Multiple cohorts/Multi-center studies
• 7.2 The role of treatment for making a prediction model
• 7.2.1 Modeling treatment
• 7.2.2 Comparative effectiveness tables
• 7.3 Learning curve paradigm
• 7.4 Internal validation (data splitting)
• 7.4.1 Single split
• 7.4.2 Calendar split
• 7.4.3 Multiple splits (cross-validation)
• 7.4.4 Dilemma of internal validation
• 7.4.5 The apparent and the .632+ estimator
• 7.4.6 Tips and tricks
• 7.5 Missing values
• 7.5.1 Missing values in the learning data
• 7.5.2 Missing values in the validation data
• 7.6 Time-varying coefficient models
• 7.7 Time-varying predictor variables
• 8 Can’t the computer just take care of all of this?
• 8.1 Zero layers of cross-validation
• 8.1.1 What may happen if you do not look at the data
• 8.1.2 Unsupervised modeling steps
• 8.1.3 Final model
• 8.2 One layer of cross-validation
• 8.2.1 Penalized regression
• 8.2.2 Supervised spline selection
• 8.3 Machine learning (two levels of cross-validation)
• 8.3.1 Random forest
• 8.3.2 Deep learning and artificial neural networks
• 8.4 The super learner
• 9 Things you might have expected in our book
• 9.1 Threshold selection for decision making
• 9.2 Number of events per variable
• 9.3 Confidence intervals for predicted probabilities
• 9.4 Models developed from case-control data
• 9.5 Hosmer-Lemeshow test
• 9.6 Backward elimination and stepwise selection
• 9.7 Rank correlation (c-index) for survival outcome
• 9.8 Integrated Brier score
• 9.9 Net reclassification index and the integrated discrimination improvement
• 9.10 Re-classification tables
• 9.11 Boxplots of rival models conditional on the outcome
• Bibliography
• Index