Comparison of Raman Spectra of Inks using Functional Data Analysis
Anjali Gupta1, James Curran1, Patrick Buzzini2, and Christopher Triggs1
1. Department of Statistics, University of Auckland
2. Department of Forensic Science, Sam Houston State University
Forensic scientists commonly use Raman spectroscopy to generate spectral data for ink samples from inkjet printers. The purpose is to examine whether the ink samples in question have originated from the same printer or not. In traditional sense, this can be thought of as an application of Hotelling’s T2 test but for very high-dimensional data. However, commonly used statistical methods fail to analyse spectral data due to its high-dimensionality and insufficient sample observations in forensic laboratories. This leads to singular variance-covariance estimators and unstable comparisons amongst the samples. In order to compare spectral data, we propose to take advantage of the work done in the field of functional data analysis and use the Hotelling’s T2 test statistic derived by Pini et al. (2018). We conduct the comparisons between cyan, magenta and yellow inks from several printers using this method. I will discuss the benefits of using functional data analysis and the results of these comparisons in this talk.
This presentation is eligible for the NZSA Student Prize.