A Multivariate Approach To Discriminating Between Batches Of Cartridge Cases Using Design Changes In The Headstamp

Abstract

A study was conducted to determine whether subtle changes could be discerned in the ICI headstamps impressed into .22 Long Rifle cartridges produced over time by the Colonial Ammunition Company, in order to define a specific period of manufacture. A large number of boxes of relevant ammunition were obtained and 117 of these were selected for sampling. An optimised set up was used to photograph the headstamps and the resultant images were processed for data collection. The data was collected in the form of Cartesian coordinates, with 22 coordinates collected from each headstamp. The coordinate data was used in a brief exploration of the data, after which the coordinates were used to calculate data transformations providing measurements of the various parameters of the headstamps. The various data transformations were assessed and a selection of these were used in the statistical analysis of the headstamps. The ability to predict the batch source of a headstamp was assessed using Linear Discriminate Analysis. The accuracy of the batch predictions were measured using a cross-validation method which showed the predictions for the dry-primed cartridges were within five to six months (on average) of the true production date and the predictions for the wet-primed cartridges were within two months (on average) of the true production date. A blind study was conducted which showed that the classification model was very accurate. An exhibit from a historic murder in New Zealand was also examined to predict its production period. Its assignation is discussed in detail with reference to the relevance of this to the murder case.