The Art of Regression Modeling in Road Safety (回归分析在道路交通安全中的应用) 英文图书。高清pdf版本。多伦多大学 道路交通安全大牛 Ezra Hauer 编纂。
该书深入浅出，形象生动，详细论述了各类数据分析方法尤其是回归模型在道路交通领域的应用，是交通运输专业学生的优秀参考书。目录如下：
Contents
1 What Is What
1. 1 Units and Their Safety Property
1. 2 Safety, Traits, and Populations
1.3 What and Are Needed for
1.4 How and Are Used: Numerical Examples
1. 4. 1 Data for Two Populations
1.4.2 Estimating E { μ } and σ { μ }
1.4.3 How Many High- μ Units Are There?
1. 4. 4 The Performance of a Screen
1.4.5 Estimating the μ of a Unit
1. 4. 6 Is the Gamma Assumption Sensible?
1. 5 The Chosen Perspective
1. 6 Summary
References
2 A Safety Performance Function for Real Populations
2. 1 The Origin
2.2 The Estimate of E { μ }
2.3 The Estimate of σ { μ }
2.4 The Two σ ’s; Homogeneity Versus Accuracy
2. 5 Summary

References
3 Exploratory Data Analysis
3. 1 Introduction
3. 2 The Data
3. 3 The Pivot Table
3. 4 Pausing for Reflection
3. 5 Visualization
3. 6 Terrain
3. 7 Summary
References
4 Curve-Fitting
4. 1 Why Do We Need to Curve-Fit?
4. 2 There is No Free Lunch
4. 3 Kernel Regression
4. 3. 1 Bandwidth and Goodness of Fit
4. 3. 2 Adding a Variable
4. 4 Summary
References
5 Preparing for Parametric Curve-Fitting: The “Solver”
5. 1 Optimization in Modeling
5. 2 Using the Solver to Find Minima and Maxima
5. 3 Solver for Curve-Fitting: An Example

5. 4 Initial Guess and Parameter Scaling
5. 5 Summary
References
6 A First Parametric SPF
6. 1 The Approach to Parametric SPF Modeling
6. 2 A Simple Parametric SPF
6. 3 Preparing and Using the First Curve-Fitting Spreadsheet
6. 4 Modifying the Objective Function
6.5 Estimating σ { μ }
6. 6 The Accuracy of Parameter Estimates
6.6.1 The Statistical Inaccuracy of β 1
6. 6. 2 The Incompleteness of “Statistical Inaccuracy”
6. 7 Regression, Design Choices, Interventions, and Safety Effect
6. 7. 1 A Road Design Example
6. 7. 2 A Speed-and-Safety Example
6. 7. 3 A Generalization
6. 7. 4 The Debate
6. 8 Summary
References
7 Which Fit Is Fitter
7. 1 Goodness of Fit
7. 2 The CURE Plot

7. 3 The Bias-in-Fit
7. 4 Leveling the Playing Field
7. 5 When Is a CURE Plot Good Enough?
7. 6 Comparing CURE Plots
7. 7 Summary
References
8 What to Optimize?
8. 1 Introduction
8. 2 Likelihood
8. 2. 1 The Parameter Behind Poisson Accident Counts
8. 2. 2 The Parameters Behind the NB Distribution
8. 3 A Few Likelihood Functions
8. 3. 1 The Poisson Likelihood Function
8. 3. 2 The Negative Binomial Likelihood Function
8. 3. 3 The Negative Multinomial Likelihood Function
8. 4 Alternative Objective Functions
8. 5 Summary
References
9 Adding Variables
9. 1 When to Add a Variable
9. 1. 1 The Necessary Conditions
9. 1. 2 The Sufficient Condition

9. 2 The Variable Introduction EDA: Is AADT Safety Related?
9. 3 How to Add a Variable to the C-F Spreadsheet
9. 4 The Omitted Variable Bias
9. 5 A Few CURE Plots
9. 6 Adding Variables: Terrain
9. 7 Panel Data and the NM Likelihood
9. 8 Panel Data and Alternative Objective Functions
9. 9 Adding Another Variable: Year
9. 10 Summary
References
10 Choosing the Function Behind the Data
10. 1 The Holy Grail
10.2 The Elusive f (): A Story with Morals
10. 3 Enroute to the Multiplicative Model Equation
10. 4 Trying for a Better Fit
10. 4. 1 Remedy I: A Bump Function for Segment Length
10. 4. 2 Remedy II: Alternative Functions
10. 5 What Equations Look Like
10. 6 Trying Various Functions
10. 7 Parameter Proliferation
10. 8 Options and Choices: Terrain Revisited
10. 8. 1 Fitting Separate SPFs

10.8.2 Making β Terrain into a Function of Other Predictor Variables
10. 9 Interaction
10. 10 Summary
References
11 Accuracies
11. 1 Considerations
11. 2 The Simulation Idea
11. 3 The Idea Executed
11. 3. 1 Determining Standard Errors
11. 3. 2 How Accuracy Is Affected by the Addition of the Terrain Variable
11. 3. 3 How Accuracy Is Affected by the AADT “Error in Variables”
11. 3. 4 Study Design
11. 4 Summary
References
12 Closure