Sample Size Considerations for Control Glass in Casework

Abstract

The variation in the refractive index of control glass samples submitted from scenes of crime was investigated. Refractive index data was compiled for four types of flat glass: toughened float glass, non-toughened float glass, non-float glass, and wired glass. From the case data, additional control glass samples for each type of glass were simulated. This was done by standardizing and pooling the data for each type of glass and then drawing a predetermined number of measurements from a given data pool. The mean, estimate of the standard deviation, and 95% and 99% confidence intervals were calculated from these larger sets of simulated, numerical data. The statistical results for each type of glass examined were tabulated and worked examples based on the data are provided. The author suggests that, as a guideline, a minimum of 17 refractive index measurements be made when sampling a toughened float glass control, and a minimum of nine refractive index measurements be made when sampling a non-toughened float glass control.

Evaluation of Match Criteria Used for the Comparison of Refractive Index of Glass Fragments*,†,‡,§

“…In other studies, Locke and co‐workers reported values of 0.00004 and 0.00005 for standard deviations of measurements of typical modern, presumably float, window glass (9,10). In more recent studies, typical reported standard deviations of RI values over sheets of various sizes are 0.00004 for nontempered float glass (7,11), 0.00007 for tempered float glass (6,11), and 0.00006 for nonfloat glass (11). The measured standard deviations shown in Table 2 are generally in agreement with those previous studies in which analytical precisions were similar to those of the method used here.…”

Forensic Science

“…The forensic significance of glass trace elemental analysis and refractive index measurements was examined for their discrimination power (141). The variation in the refractive index of control glass samples submitted from crime scenes was investigated (142). The robustness of the Bayesian approach to forensic glass analysis has been demonstrated (143).…”