© 2004 by Oxford University Press
Multivariate analyses of materials found on a sentenced man and on the scene of the crime
1 Institute of Biology, Department of Botany and Plant Physiology, University of Oslo, Norway, 2 Institute of Biology, Laboratory of Electron Microscopy, University of Oslo, Norway, 3 Betzy Kjelsbergs vei 5 A, 0486 Oslo, Norway
In 1957 a girl was brutally murdered in Oslo, Norway, and a 23 year old man, who claimed to be innocent, was sentenced for the crime. To examine that claim, materials collected at various locations on the scene of the crime and on the garments of the sentenced man have been reanalysed using two multivariate techniques, namely cluster analysis and ordination by detrended correspondence analysis (DCA). The cluster analysis divided the different locations in three: one group with locations from the sentenced man, one group for related locations from the scene of the crime, and finally locations with little association with other locations. The main axis of the DCA largely separated the locations so that those on the sentenced man had a low loading while locations on the scene of the crime had a high loading; there was little overlap between the two sets of locations. In further simulations materials found on the scene of the crime were scored for being present on a shoe belonging to the sentenced man. This procedure increased the loading of the shoe on the main axis and drew that location away from the other locations on the sentenced man and towards those at the scene of the crime. This simulation thus suggests that DCA may be a sensitive method for unravelling a possible association between a suspect and the scene of the crime. This study shows no association between the sentenced man and the scene of the crime.
Keywords: classification; cluster analysis; correspondence analysis; detrended correspondence analysis; forensic science; multivariate analyses; ordination; reciprocal averaging
Received 20 January 2003. Revised 8 November 2004. Accepted 11 November 2004.