Online Program

Optimization of Headspace Sampling Parameters for Forensic Fire Debris Analysis with Discrete Outcomes (309909)

Keywords: Discrete data, generalized linear models, forensic science, optimization, response surface methodology

Dynamic headspace sampling offers several potential advantages in extracting ignitable liquid (IL) residue from fire debris compared to the standard method of static headspace sampling. However, optimum levels for the sampling parameters temperature and collection volume (CV) must be established for this technique to be adoptable by the forensic science community. In this work, researchers dynamically sampled the headspace vapor of an IL containing a known number of recoverable target compounds for different levels of temperature and CV, recording the number of compounds identified in the sample and the number of unique compounds identified in the breakthrough vial. Compounds in the breakthrough vial represent lost IL signal, so the desired optimal settings are those which maximize the number of identified compounds while minimizing the number captured in breakthrough. Analysis of this data is complicated by potential correlation between these two quantities. We apply generalized linear models and response surface methodology to model these compounds, obtaining an estimate of the range of settings that result in optimal sample processing. Through simulation, we explore how the discrete nature of the data affects the uncertainty of the optimum obtained from the fitted response surface.