Assessing evidentiary value in fire debris analysis by chemometric and likelihood ratio approaches

Results are presented from support vector machine (SVM), linear and quadratic discriminant analysis (LDA and QDA) and k-nearest neighbors (kNN) methods of binary classification of fire debris samples as positive or negative for ignitable liquid residue. Training samples were prepared by computationally mixing data from ignitable liquid and substrate pyrolysis databases. Validation was performed on an unseen set of computationally mixed (in silico) data and on fire debris from large-scale research burns. The probabilities of class membership were calculated using an uninformative (equal) prior and a likelihood ratio was calculated from the resulting class membership probabilities. The SVM method demonstrated a high discrimination, low error rate and good calibration for the in silico validation data; however, the performance decreased significantly for the fire debris validation data, as indicated by a significant increase in the error rate and decrease in the calibration. The QDA and kNN methods showed similar performance trends. The LDA method gave poorer discrimination, higher error rates and slightly poorer calibration for the in silico validation data; however the performance did not deteriorate for the fire debris validation data.