NCJ Number
224719
Journal
Journal of Forensic Sciences Volume: 53 Issue: 5 Dated: September 2008 Pages: 1092-1101
Date Published
September 2008
Length
10 pages
Annotation
This study examined the variation in gasoline components in 20 retained liquid gasoline samples from fire investigations in the New York City area using gas chromatography coupled with a mass spectrometer (GC-MS).
Abstract
The study found that the variation in the sample population was sufficient to distinguish all the samples from one another using principal component analysis (PCA), canonical variate analysis (CVA), and orthogonal canonical variate analysis (CVA). CVA differentiated all samples in the population using three dimensions; OCVA required four dimensions. These results were cross-validated by using the “jackknife” method to confirm the classification functions. By plotting the CVA and OCVA data in two and three dimensions, clearly defined and easily interpretable clusters were evident in the sample population. PCA required at least 10 dimensions of data in order to predict the correct groupings. Preliminary studies conducted on weathered gasoline samples showed that group predictions using CVA and OCVA were applicable to approximately 75 percent to 80 percent weathered by volume. A more formal and detailed study will be conducted with these weathered samples. The outcome of this initial study developed the multivariate procedures and methods for adaptation to planned future studies. The hope is that future studies will produce practical procedures that have the scientific rigor required of a technique applicable to fire debris casework. The statistical methods used are described in detail. 7 tables, 6 figures, and 27 references