Study guide

Regression to the Mean and Empirical Bayes Interpretation

Road Safety Professional RSP

Track your progress as you read

Finished sections get a check mark.

1 Exam Orientation and Preparation

RSP1 Exam, Eligibility, and Certification Preparing Across the Five RSP1 Domains

2 Foundations: Definitions, Evidence, and Standards

Road Safety Definitions and the Public Health Burden Evaluating Road Safety Literature and Information Nominal and Substantive Safety: Standards and Performance

3 Foundations: Contributing Factors, Energy, and Users

Interacting Crash Contributing Factors Collision Severity, Speed, and Protective Layers Road User Characteristics, Abilities, and Vulnerability

4 Foundations: Partners and Safety Frameworks

Safety Partners, Responsibilities, and Multidisciplinary Perspectives Haddon Matrix and the Development of Safety Management Safe System Principles and Protective Layers Vision Zero, Toward Zero Deaths, and Safety Culture

5 Measurement: Data Collection and Quality

Safety Data Types, Sources, and Users Collecting Quantitative and Qualitative Safety Data The Six Safety Data Quality Attributes Data Process Reviews, Governance, and Investment Decisions

6 Measurement: Exposure, Severity, Models, and Effects

Crash Frequency, Exposure, and Rates Injury Severity, Crash Costs, and Weighted Measures Safety Performance Functions and Suitable Models Regression to the Mean and Empirical Bayes Interpretation CMF Definitions, Scope, and Evidence Quality

7 Human Factors: Perception and the Driving Task

Sensory Perception, Attention, and Workload Driving Task and Perception-Response Time

8 Human Factors: Behavior and Programs

Impairment, Distraction, Fatigue, and Speed Choice Enforcement, Licensing, and Behavioral Countermeasures Outreach, Social Norms, and Audience-Centered Programs

9 Human Factors: Infrastructure, Multimodal Design, and Guidance

Roadway Features, Expectancy, and Behavioral Adaptation Human Factors in Multimodal Planning, Design, and Operations Positive Guidance and Driver Expectancy

10 Problem Solving: Management, Screening, and Diagnosis

The Road Safety Management Process Site-Specific and Systemic Network Screening Crash Patterns, Collision Diagrams, and Diagnostic Hypotheses Field Investigation and Surrogate Safety Measures

11 Problem Solving: Countermeasures and Safety Audits

Countermeasure Sources, Mechanisms, and Selection Combining CMFs and Evaluating Treatment Trade-Offs Road Safety Audits and Documented Owner Responses

12 Problem Solving: Economics and Project Selection

Economic Appraisal: Benefits, Costs, and Discounting Prioritizing Projects and Nonmonetized Impacts

13 Problem Solving: Evaluation Methods and Planning

Evaluating Countermeasures: Before–After and Cross-Sectional Studies Planning Evaluations and Collecting Credible Evidence

14 Implementation: Statewide and Local Plans

Developing Strategic Highway Safety Plans Local, Tribal, and Regional Safety Action Plans

15 Implementation: Funding, Policies, and Practicality

Infrastructure Safety Funding and Program Eligibility Safety Policies, Behavioral Programs, and Practicality

16 Implementation: Teams, Communication, and Learning

Teams, Champions, Leadership, and Organizational Safety Culture Communication, Public Support, and Implementation Consensus Program Monitoring, Safety Targets, and Plan Updates

Regression to the Mean and Empirical Bayes Interpretation

Key Takeaways Explain This Topic AI

RTM is a tendency, not a guaranteed next-period decline.

Use the same period and outcome for predicted, observed, and expected counts.

The low-observation example gives an expected count of 2.49; the three-year example gives 18.27 annually.

Positive or negative excess is a screening comparison, not proof of treatment potential or safety.

Last updated: October 2026

  • Regression to the Mean and Empirical Bayes Interpretation
  • Observations fluctuate around underlying performance

Crash counts vary from period to period even when exposure and conditions are stable. A site selected because of an unusually high recent count may have a lower count later without treatment. This regression-to-the-mean (RTM) tendency complicates screening and before-after evaluation. It does not guarantee a decline at every site or prove that a treatment had no effect.

Consider a curve with a long-run average near 1.5 crashes per year and a recent year with seven. A later average near 1.8 after signs were installed cannot identify how much change came from signs versus random variation or other conditions. The original selection based on the spike creates a particular risk of overcrediting the treatment.

Combine model information and site history

Empirical Bayes (EB) combines an applicable prediction with observed history to estimate long-term expected frequency. Model information reflects comparable sites and conditions; observed history adds site-specific information. The method reduces RTM bias when its assumptions, model, and data are suitable. It does not make uncertainty disappear or identify the physical cause of every excess crash.

For a simplified common formulation using cumulative counts over the same study period : E = w P + ( 1 − w ) O E=wP+(1-w)O E = wP + ( 1 − w ) O

P P P is the summed predicted count, O O O is the observed count, and E E E is the EB-expected count. A common weight is: w = 1 1 + k P w=\frac{1}{1+kP} w = 1 + k P 1 ​

k k k is the appropriate dispersion parameter under the model's formulation. Other procedures can specify different details; use the documented method. Do not mix an annual prediction with a multi-year observed total or reuse a dispersion parameter with incompatible units.

Interpret the weight

The weight on the model is w w w , and the weight on observations is 1 − w 1-w 1 − w . With a larger product k P kP k P , the observed history receives more weight. With a smaller product, the model receives more weight. The formula does not directly use how extreme the observed count is to choose w w w in this simplified version.

A longer period can increase cumulative predicted counts and shift weight toward the observed history, holding other assumptions constant. This does not prove the observations are free of reporting errors or that the model is irrelevant. Data quality remains essential. A zero recorded count with incomplete capture should not be treated as reliable evidence of zero risk.

Check three worked examples

The following figures are illustrative and use the cumulative-count formulation above.

  • Case Observed count O Predicted count P k Weight w Expected count E
  • One-year high observation 6 0.8 0.50 0.7143 2.29
  • One-year low observation 1 5.2 0.35 0.3546 2.49
  • Three-year record 54 72 0.30 0.04425 54.80

For the first case, w = 1 / ( 1 + 0.5 × 0.8 ) = 0.7143 w=1/(1+0.5\times0.8)=0.7143 w = 1/ ( 1 + 0.5 × 0.8 ) = 0.7143 and E = 0.7143 × 0.8 + 0.2857 × 6 = 2.2857 E=0.7143\times0.8+0.2857\times6=2.2857 E = 0.7143 × 0.8 + 0.2857 × 6 = 2.2857 . For the second, w = 1 / ( 1 + 0.35 × 5.2 ) = 0.3546 w=1/(1+0.35\times5.2)=0.3546 w = 1/ ( 1 + 0.35 × 5.2 ) = 0.3546 and E E E is about 2.49 . For the third, use the full three-year prediction of 72; the expected annual mean is 54.7965 / 3 54.7965/3 54.7965/3 , about 18.27 . Keep unrounded intermediate values when comparing totals.

Expected counts are estimates of means and can be fractional. They are not a prediction that a fraction of a crash will literally occur. This distinction helps explain why integer observations and decimal model estimates coexist.

Interpret excess expected frequency

An excess-expected measure compares EB-expected frequency with a suitable predicted peer baseline. If both are annual means, a simplified Potential for Safety Improvement (PSI) expression is E − P E-P E − P in annual units. If using cumulative counts, both terms must cover the same period. Some screening definitions distinguish base-condition and site-condition predictions; follow the selected method's documentation.

A positive value identifies performance above the modeled comparison and may support priority for diagnosis. It does not prove that all the excess is treatable or caused by geometry. A negative value does not prove the site is safe or that no systemic treatment can help. The comparison inherits model and data limitations.

Recognize assumptions and uncertainty

EB needs an appropriate SPF, compatible crash definitions, relevant exposure, and a dispersion parameter suitable for the model. A model from the wrong facility type can pull an estimate toward an unsuitable peer mean. Incorrect location or severity coding can distort both history and prediction.

The method addresses a statistical problem; it does not replace field diagnosis or measure every contributing factor. Interpret the result alongside collision patterns, users, roadway conditions, and feasible treatment evidence. FHWA network-screening guidance provides context for selecting suitable performance measures.

Use the estimate in the management process

A screening estimate can identify where further investigation is valuable. A before-after evaluation can use EB to estimate what would have occurred without treatment, with after-period exposure and other relevant changes incorporated. These are related applications, but an expected-before estimate alone is not a completed treatment-effect evaluation.

For a nontechnical audience, explain that the estimate combines site experience with comparable-road evidence to reduce overreaction to an unusual year. Report uncertainty and the next diagnostic step. Avoid promising that a sophisticated method “eliminates every bias” or proves a project's future benefit.