Evaluation of Chan’s tetrahedron classification method and Karweng dissimilarity index using heroin impurity data

Statistical evaluation becomes increasingly important in forensic profiling as there is no single statistical method available to cater to all types of data. In this preliminary study, 20 heroin samples were used as linked samples and non-linked samples to evaluate a recently developed classification method called Chan’s tetrahedron model. Four commonly found impurity peaks were chosen for use to assess the Karweng dissimilarity index derived from this model. Throughout the study, the performance of three basic configurations and the associated dissimilarity indices obtained from the tetrahedron method were compared with that of the established statistical techniques. The obtained tetrahedrons illustrated comparable graphical outcomes with the score plot and dendrogram respectively obtained from principal component analysis (PCA) and hierarchical cluster analysis (HCA). Histograms were constructed and the Karweng dissimilarity index displayed significantly better levels of errors (fewer false results) than modified Pearson and cosine. In addition, the index also proved to be able to show excellent discriminating power and the best area under the receiver operating characteristic (ROC) curve. This study infers that this new classification method could be as useful as the conventional classification techniques. However further evaluation using larger datasets is recommended.