Applied Statistical Considerations for Clinical Researchers

Preface
Acknowledgements
Contents
Chapter 1: Introduction
1.1 Outline of Chapters
References
Chapter 2: Preliminaries
2.1 Testing Where You Are Now …
2.2 Refreshing Existing Knowledge; Where to Go, How to Do It
References
Chapter 3: Design
3.1 Design from Different Viewpoints
3.1.1 Design from the Clinician’s Perspective
3.1.2 Design from the Statistician’s Perspective
3.1.3 An Example of the Provision of Support in Research Design
3.1.4 Design from the Perspectives of Patients and of the Public
3.1.5 Design from the Perspective of the Funder
3.2 Why Design Research Studies?
3.2.1 Bias
3.2.2 Precision
3.3 Types of Design
3.3.1 Basic Designs
3.3.1.1 Experimental Design—Parallel Group
3.3.1.2 What If We Cannot Randomise?
3.3.1.3 Observational Design—Cohort Study
3.3.1.4 Observational Design—Case-Control Study
3.4 Design and Evidence
3.4.1 The Pyramid (and Levels) of Evidence
3.4.2 Conduct Considerations
3.5 More Complexity in Design
3.5.1 Clustering and Hierarchical Structures
3.5.2 Some Other Experimental Designs
3.5.3 Sampling Schemes
3.5.4 Missing Data
3.5.5 Studies of a Different Nature
3.5.5.1 Diagnostic Accuracy Studies
3.5.5.2 Method Comparison Studies
3.6 Some Examples of Good Design
3.6.1 Example 1 (Randomised Controlled Trial)
3.6.2 Example 2 (Cohort)
3.6.3 Example 3 (Case-Control)
3.7 Constraints and Compromises
References
Chapter 4: Planning
4.1 Project Planning in General
4.2 The Study Protocol
4.3 Planning for Data
4.4 Planning for Analysis
4.4.1 The Statistical Analysis Plan (SAP)
4.4.2 Planning for Sample Size
4.5 Planning for Reporting and Dissemination
References
Chapter 5: Data I
5.1 Data Acquisition
5.2 File Type, Format, and Other Properties
5.2.1 Data Format
5.2.2 Storage
5.2.3 Speed
5.2.4 Security
5.3 Typical Data Sources and Structures
5.3.1 Trials Data
5.3.2 Other Primary Data
5.3.3 Secondary Data—Electronic Health Records
5.3.4 Survey and Questionnaire Data
5.4 Pre-processing: Linking, Joining, Aggregating
5.5 The ‘Burden’ of Data and Its Consequences
References
Chapter 6: Data II
6.1 Restructuring from Raw Data
6.1.1 Scenario 1: Merging Individual Participant Datasets
6.1.2 Scenario 2: Tackling EHRs with Repeated Events and Other Complexities
6.1.3 More Complicated Restructuring
6.2 Variables
6.2.1 What’s in a Variable?
6.2.2 Types of Variable—A Refresher
6.2.3 Numeric Variables
6.2.3.1 Discrete Variables
6.2.3.2 Continuous Variables
6.2.4 Categorical Variables
6.2.4.1 Nominal Variables
6.2.4.2 Ordinal Variables
6.2.4.3 Tips on Categorisation
6.2.5 Date Variables
6.3 Data Cleaning
6.3.1 Correcting and Amending Data
6.3.2 Format Checking
6.3.3 Range Checking and Related Verification
6.3.4 Cross-Checking
6.4 Reorganising Data Structures
6.4.1 Manipulating Data to Create New Variables
6.4.1.1 Variable-Level Tasks
6.4.1.2 Dataset-Level Tasks
6.4.1.3 Filtering
6.4.2 Relational Databases
6.5 Other Routine Data Issues
6.5.1 Backups
6.5.2 Audit Trails
6.5.3 Security and Storage
6.5.4 Data Encryption
References
Chapter 7: Analysis
7.1 Some Prerequisites
7.1.1 Summary Statistics
7.1.2 Statistical Distributions
7.1.3 Hypothesis Tests; Statistical Significance; p-Values
7.1.4 Confidence Intervals
7.1.5 The Outcome Variable
7.1.6 The Explanatory Variables
7.2 Descriptive Analysis
7.2.1 Structuring Descriptions
7.2.2 Statistical Software
7.2.3 Pictures—Plotting Your Data
7.2.3.1 Continuous Variables
7.2.3.2 Discrete Variables
7.2.3.3 Categorical Variables
7.2.3.4 Plots with More Than One Variable
7.2.3.5 Plot Types Depend on Distributional Shape
7.2.4 Numerical Summary Statistics
7.2.5 Table 1—Your Data Described
7.3 Statistical Tests
7.3.1 Tests for Continuous Outcomes
7.3.1.1 The t Test
7.3.1.2 The Mann-Whitney U Test
7.3.1.3 Three or More Groups
7.3.1.4 The Kruskal-Wallis Test
7.3.2 Tests for Categorical Outcomes
7.3.2.1 A Brief Aside: What Is a ‘Test Statistic’?
7.3.2.2 Conducting a Basic Chi-squared Test
7.3.3 Structural Dependence—Paired and Clustered Data
7.3.3.1 More Complex Examples of Clustering
7.4 Sample Size
7.4.1 The Basic Formula
7.5 When a Single Test Is Not Enough
References
Chapter 8: Regression
8.1 Linear Regression—A Brief Recap
8.1.1 Multiple Linear Regression
8.1.2 Flexibility of the ‘Right-Hand Side’
8.2 Linearity
8.2.1 Why Linear?
8.2.2 The Anscombe Quartet
8.3 Types of Regression Model
8.3.1 Continuous Outcome: Linear Regression
8.3.1.1 Multiple Linear Regression
8.3.1.2 Example: A Hypothetical Study on Weight Gain/Loss
8.3.1.3 Interpretation
8.3.1.4 Assumptions
8.3.1.5 Diagnostics
8.3.2 Binary Outcome: Logistic Regression
8.3.2.1 The ‘Outcome’ in Logistic Regression
8.3.2.2 Interpretation
8.3.2.3 Assumptions
8.3.2.4 Diagnostics and Model Fit
8.3.3 Counts and Rates: Poisson Regression
8.3.3.1 Count Variable as Outcome
8.3.3.2 Rate as Outcome
8.3.3.3 Further Notes on Counts and Rates
8.3.4 Other Types of Linear Regression
8.4 Reasons for Using Regression
8.4.1 Exploratory Associations
8.4.2 Estimation of Effects
8.4.2.1 Methods of Entering Variables
8.4.3 Prediction
References
Chapter 9: Complexity
9.1 Dependence, Clustering and Hierarchy
9.1.1 Matching in Regression Models
9.1.1.1 Analytical Methods
9.1.2 Clustering and Hierarchy
9.1.3 Allowing for Repeated Measures
9.2 More Complex Regression Models
9.2.1 Time-to-Event Outcome (Survival Analysis)
9.2.1.1 Tests and Plots
9.2.1.2 Cox Regression and the Proportional Hazards Model
9.2.1.3 Parametric Models and Other Frameworks
9.2.2 Ordinal Outcome: Proportional Odds Model
9.2.3 Other Regression Models
9.2.3.1 Fractional Polynomials
9.2.3.2 Quantile Regression
9.2.3.3 Time Series and Auto-Regression
9.2.3.4 Multivariate Regression
9.3 Missing Data
9.3.1 Types of Missing Data
9.3.2 Options Available
9.3.3 Multiple Imputation
9.3.4 In Practice
References
Chapter 10: Inference
10.1 What Can We Infer?
10.1.1 Randomised Controlled Trials
10.1.2 Observational Studies
10.1.3 Appropriateness
10.1.4 P-Values
10.1.5 Confidence Intervals
10.2 Causality
10.2.1 The Bradford Hill Criteria for Causality
10.3 Assumptions
10.3.1 Sampling and the Study Setting
10.3.2 Participation Bias
10.4 Multiplicity
10.4.1 Observational Studies
References
Chapter 11: Dissemination
11.1 General Issues
11.2 Channels and Delivery Types
11.2.1 Written
11.2.1.1 High Impact Reports
11.2.1.2 Peer-Reviewed Journal Articles
11.2.1.3 Tables
11.2.1.4 P-Values
11.2.1.5 Figures
11.2.1.6 Main Body Text
11.2.1.7 Conference Posters
11.2.2 Spoken
11.2.2.1 Conference Presentations
11.2.2.2 Seminars/Webinars
11.2.3 Discussion
11.2.4 Other Channels
11.2.4.1 Traditional Media
11.2.4.2 Social Media
11.3 Patients and the Public
11.4 Resources
References
Chapter 12: A Conversation
12.1 Setting the Scene
12.2 Meeting Up
12.3 Data
12.4 Analysis
12.5 Preparing to Disseminate
12.6 Presentation and Manuscript
12.7 Finishing Off
12.8 Post Script
Reference
Chapter 13: Conclusions
13.1 Some Topics Not Covered
13.1.1 Statistical Genetics
13.1.2 Pre-clinical Studies
13.1.3 Diagnostic and Prognostic Studies
13.1.4 Health Economics
13.1.5 Time Series Analysis
13.1.6 Meta-Analysis
13.1.7 Machine Learning
13.1.8 Bayesian Inference
13.2 Where to Now?
Appendix
Solutions to the Quiz
Scoring Scheme
Explanatory Notes for the Questions and Answers
Index