Next >

Handbook of Forensic Statistics

Section I: Perspectives on Forensic Statistics The History of Forensic Inference and Statistics: A Thematic Perspective Introduction Forensic Science and the Evaluation of Evidence The Need for an Interpretative Model Support of Judicial Disciplines for a Scientific Presentation of the Value of Evidence Probability of Proposition Given Evidence and of Evidence Given Proposition Quantification of the Value of Evidence Using Alternative Numerical Summaries Change from Two-Stage Approach to Continuous Approach Presentation of Evidence: New Challenges to Solve The Island Problem and Results of a Database Selection Profile Probability vs Conditional Profile Probability Evaluation by Taking Errors into Account A Minimum Value for the Profile Probability Propositions and Pre-Assessment The Choice of Propositions The Pre-Assessment Translation of a Numerical Value into a Verbal Equivalent Assessment of Performance Role for Likelihood Ratio as aMeasure for Investigation as Well as for Evaluation Probabilistic GraphicalModels Bayesian Networks Bayesian Networks to Manage ‘Masses’ of Evidence Bayesian Networks in Judicial Contexts Bayesian Networks in Forensic Science: Particular Case Modeling Bayesian Networks in Forensic Science: Generic Patterns of Inference Not Only Inference: The Way to Make a Decision The Objectives and Ingredients of Decision Theory Graphical Models The Existence or Otherwise of a True Value of the Evidence Acknowledgments References Section II: General Concepts andMethods Frequentist Methods for Statistical Inference Introduction Definitions and Notation Data and Evidence Random Variables and Probability Distributions Sampling from a Distribution or Population Estimation Properties of Point Estimators Estimating Allele Proportions A Point Estimate Constructing a Confidence Interval Choosing a Confidence Coefficient Estimating a False Positive Probability Through an Experiment The Design of Experiments to Test Categorical Source An Experiment to Test Categorical Judgments of Latent Print Examiners Constructing Confidence Intervals Interpreting Confidence Intervals p-Values p-Values in a Comparison of Glass Fragments Interpreting p-Values Hypothesis Tests Classical Hypothesis Tests for Refractive Index Matching Type I Errors and the Size of a Test Type II Errors and the Power of a Test Hypothesis Testing with p-Values Hypothesis Testing with Confidence Intervals Issues in Interpreting the Results of Hypothesis Tests, p-Values, and Confidence Coefficients Transposition Multiple Tests: Proof of the Null Hypothesis and Adjusted p-Values Arbitrary Lines Alternatives and Likelihoods Resampling Methods Bootstrap Estimates Permutation Tests Acknowledgments References Bayesian Methods and Forensic Inference Introduction The Basics A Beta-Binomial Mock Example A Gamma-Poisson Mock Example Markov Chain Monte Carlo Broad Applications Summary Acknowledgments References Comparing Philosophies of Statistical Inference Inferential Philosophies Frequentist Inference Bayesian Inference Other Approaches to Inference Fiducial Inference Likelihood Inference Confidence Distributions Comparing the Approaches Planning Studies Using Frequentist Inference Challenges for Frequentist Inference Flexible Inference with Bayesian Methods Model Modifications and Adjustments The Prior Distribution and the Definition of Probability Relevance to Forensic Statistics Likelihood Ratios and Bayes Factors Two-Stage Procedures in Forensic Science Forensic Evidence as Expert Opinion and Error Rates Summary References Decision Theory Introduction Concepts of Statistical Decision Theory Preliminaries: Basic Elements of Decision Problems Utility Theory Implications of the Expected Utility Maximisation Principle The Loss Function Particular Forms of the Expected Utility Maximisation Principle Likelihood Ratios in the Decision Framework Decision Theory in the Law and Forensic Science Legal Applications Forensic Science Applications Forensic Identification Understanding Probability Assignment as a Decision: The Use of Proper Scoring Rules Other Forensic Decision Problems: Consignment Inspection Discussion and Conclusions Further Readings Forensic Science General Acknowledgments References Association Does Not Imply Discrimination: Clarifying When Matches Are (and Are Not) Meaningful Introduction Association and Discrimination Quality of Test: Sensitivity and Specificity Sources of Error Weight of Evidence: The Likelihood Ratio Useful Databases for Ascertaining Discriminatory Power Conflating Conditional Statements: The Prosecutor’s Fallacy Examples: The Discriminatory Power of Forensic Evidence Arson Investigation Other Types of Forensic Evidence: DNA, Fingerprints, and Shoe Prints Abusive Head Trauma Conclusion References Validation of Forensic Automatic Likelihood Ratio Methods Introduction Scope Aim Structure Validation Process Standardization Validation of Theoretical and Empirical Aspects Performance Characteristics for Automatic LR Methods Empirical Validation Validation Protocol Primary Performance Characteristics Performance of Probabilities by Proper Scoring Rules Discrimination and Calibration of Probabilities Performance of Likelihood Ratios Properties of Well-Calibrated Likelihood Ratios Examples with Primary Performance Characteristics Secondary Performance Characteristics Robustness Monotonicity Generalization Conclusion References Bayesian Networks in Forensic Science Introduction Probability Logic Simple Bayesian Networks for Forensic Problems Object-Oriented Bayesian Networks Forensic Genetics Bayesian Networks for Simple Criminal Identification Simple Disputed Paternity Bayesian Networks for Complex Criminal Cases Involving Family Relationships Mutation Bayesian Networks for Analysing Mixed DNA Profiles Discrete Features Continuous Features Analysis of Sensitivity to Assumptions on Founder Genes Uncertainty in Allele Frequencies Heterogeneous Reference Population Conclusions Appendix 8A: Bayesian Network Basics A.1 Qualitative Structure A.2 Independence Properties A.3 Quantitative Structure A.4 Computation References Section III: Legal and Psychological Dimensions How Well Do Lay People Comprehend Statistical Statements from Forensic Scientists?Methodological Overview Consistency Framing Format Sensitivity (In)Coherence Prosecutor’s Fallacy Defense Attorney’s Fallacy Directional Errors Aggregation Errors Ability Orthodoxy Discussion Conclusion References Forensic Statistics in the Courtroom The Purpose, Form, and Prerequisites of Expert Testimony Lay and Expert Testimony Qualifications for Statistical Experts (and Experts Who Use Statistics)Forms of Statistical Expert Testimony Reasonable Scientific or Statistical Certainty Special Rules for Scientific Expert Testimony*The General-Acceptance Standard The Scientific-Validity Standard Selected Evidentiary Issues in Forensic Statistics Two Uses of Statistical Analysis as Evidence Theory and Application Error Rates in Determining Admissibility Error Rates, Likelihood Ratios, and Bayes Factors for Quantifying Probative Value Conclusion References Cases and Rules Section IV: Applications of Statistics to Particular Fields in Forensic Science DNA Frequencies and Probabilities Introduction Likelihood Ratios Population Genetics Single Loci Multiple Loci Population Structure Lineage Markers Mixtures Semi-Continuous Model Continuous Model DNA Sequence Data Future Directions Conclusions References Kinship Introduction Genetic Models for Allele and Genotype Frequencies Population Structure Relatedness Relatedness and Population Structure Parentage Calculations for Structured Populations Identifying Remains SNP Data Genealogy Data Conclusions References Statistical Support for Conclusions in Fingerprint Examinations Introduction Fingerprint Examination Framework (ACE-V)Analysis Stage Comparison Stage Evaluation Stage Verification Stage Decision Making in the ACE-V Framework Summary Common Source vs. Specific Source Scenarios Common Source Scenario Specific Source Scenario Simulations Summary Weight of Fingerprint Evidence Similarity Metrics and Kernel Functions Score-Based Likelihood Ratios Common Source Score-based Models Suspect-Centred Score-based Models Trace-Centred Score-based Models Approximate Bayes Factor for Fingerprint Evidence Summary Factors Affecting Examiners’ Decision Making and Error Rates Decision Making During ACE PCAST Controversy Statistical Analysis of Experiments Designed to Study the ACE-V Framework Summary U.S. Defense Forensic Science Center’s Summary Conclusions References Probabilistic Considerations When Interpreting Database Search and Selection Effects Introduction Probabilistic Analysis Scientific Debate: The Database Search Controversy The Number of Adventitious Matches Reporting DatabaseMatches Comparing a “Probable Cause” and a “Database Search” Match The Other Evidence in the Case Guidance on Interpreting Database Matches Dealing with “Selection Effects”More Generally Interesting Cases Conclusion References Comparing Handwriting in Questioned Documents Introduction Outline of This Chapter Historical Background Roman Law English Law United States Law Challenges in Studying Questioned Documents The Art of Conducting a Handwriting Examination Principles of Handwriting Analysis The ACE-V Process of Comparing Handwriting A Probability Scale for Expressing Conclusions Statistical Approaches to Handwriting Analysis Comparing Howland Will Signatures A Bayesian Approach to Comparing Signatures A Dynamic Model for Handwriting Proficiency Studies and Error Rates Discussion Acknowledgments References Cases and Statutes An Introduction to Firearms Examination for Researchers in Statistics Introduction The Anatomy of Guns and of Ammunition A Brief History of Firearms Examination Microscopic Imperfections Comparison of Cartridge Case Marks Comparison ofMarks on Land Engraved Areas of Bullets Pairwise Comparisons From Land-to-Land Comparisons to Bullet-to-Bullet Comparisons Results for the Hamby Set of Bullets Revisiting the Question of Source Some Final Thoughts References Shoeprints: The Path from Practice to Science Introduction The Process of Evaluating Shoeprints The Challenges that Arise from Analyzing the Working Procedure Requirements for Good Practice How to Strengthen the Scientific Foundation for Footwear Analysis?Wishlist for and Design of the Semi-Automated ElementAccidentalSensor An Early Stage Implementation of the Semi-Automated ElementAccidentalSensor Preprocessing and Binarization Element Detection Detecting Accidentals and Wear Discussion and Conclusions Appendix 17A A.1 Shoeprint Comparison Stages A.2 The Conclusion Scale References Forensic Glass Evidence Introduction Early Cases of Forensic Glass Evidence Types of Glass Glass Breakage and Recovery Physical and Optical Properties Chemical Composition Methods forMeasuring Glass Elemental Concentrations ICP-MS and LA-ICP-MS Micro X-Ray Fluorescence (XRF)Multivariate Methods Glass Data Data Sets Variability Validating Procedures Simulation Strategy Simulated Match Rates Another Interpretation of “Difference”Conclusions References Estimation of Insect Age for Assessing Minimum Post-Mortem Interval in Forensic Entomology Casework Background Isomorphen and Isomegalen Diagrams Thermal Summation Models Curvilinear Models SpectralMeasurements Growth Curve Reconstruction Conclusion References Statistical Models in Forensic Voice Comparison Introduction Feature Extraction Mel-Frequency Cepstral Coefficients (MFCCs)Deltas and Double Deltas Voice-Activity Detection (VAD) and Diarization Mismatch Compensation in the Feature Domain Cepstral-Mean Subtraction (CMS) and Cepstral-Mean-and-Variance Normalization (CMVN)Feature Warping GMM-UBM Training the Relevant-Population Model (the UBM): Expectation Maximization (EM) Algorithm Training the Known-Speaker Model: Maximum a Posteriori (MAP) Adaptation Calculating a Score Remarks Regarding UBM Training Data i-Vector PLDA i-Vectors i-Vector Domain Mismatch Compensation (LDA)PLDA DNN-Based Systems DNN Senone Posterior i-Vector Systems Bottleneck-Feature Based Systems DNN Speaker Embedding Systems (x-Vector Systems)Score-to-Likelihood-Ratio Conversion (Calibration)Validation List of Published Validation Studies Conclusion Acknowledgments Appendix 20A: Mathematical Details of T Matrix Training and i-Vector Extraction References Legal References Bringing New Statistical Approaches to Eyewitness Evidence Introduction The Eyewitness Task System and Estimator Variables Current Statistical Methodologies Diagnosticity Ratio, Discriminability Index, and ROC Curves Calibration Curve ROC Curves Logistic Regression Expected Utility Statistical Models From Diagnostic Medicine Logitnormal Bivariate Random-Effects Model Nonparametric Meta-Analysis for Diagnostic Accuracy Studies Supervised Learning Classification Methods Machine Learning Classification Models Graphical Models Tools Based on ROC Methods Methodology Development Predictive Receiver Operating Characteristic Curve Multivariate ROC Curves AUC Estimation Confidence Intervals for ROC Estimating Probability of Eyewitness Accuracy Probability of Accuracy Example Discussion Acknowledgements References

Next >