作者: Donald A. Berry
关键词: Sample (material) 、 Forensic identification 、 Statistics 、 Population 、 Mathematics 、 Suspect 、 Likelihood function 、 Bayes factor 、 Bayes' theorem 、 Density estimation
摘要: Forensic laboratories use lengths of fragments from several locations human DNA to decide whether a sample body fluid left at the scene crime came suspect or recovered suspect's clothing is victim's. Using an inferential approach called "match/binning," they first there match between and samples. If match, then calculate "match proportion." This proportion data base fragment that would similarly is, occur in interval "bin" containing length sample. Match/binning reasonable method scientific setting, other settings allow for flexibility, but it has characteristics make undesirable courts. One based on yes/no decision: arbitrary cut-off point some deemed not can be arbitrarily close others do match. Another same applies suspects whose just barely corresponding as perfectly. article describes alternative approach, one criterion. The distribution laboratory's measure- ment errors used infer form likelihood function. Then ratio guilt innocence calculated Bayes' theorem applied. focus this contribution evidence probability guilty. An important step estimating population lengths, attemp- ting account both laboratory measurement error sampling variability. two approaches are compared actual murder case (New York v. Castro). Applying criterion literally re- sulted exclusion, its scientists claimed was very small. shows correct conclusion far less clear. profiling also useful inferring parentage, example cases disputed paternity. allows calculating alleged father true father.