Abstract by Hyejung Lee
Detection of Analytes with Bivariate Absorbances in Liquid Chromatography
Liquid chromatography (LC) is one of the most common methods of analysis of chemical compounds in a sample. A simple, portable absorbance detector that has recently been developed at BYU provides absorbances at two different light wave lengths. I present the statistical methods for analyzing elution spectra for this portable machine by addressing two fundamental issues: removal of background signal and deconvolving target compounds that have overlapping elution patterns. Empirically, the form of the eluting peaks are skewed-normal shape, rather than simply a Gaussian shape. Estimating the multiple parameters of a mmixture of multiple analytes entails an iterative technique. Here, I present the use of an EM type algorithm to estimate the parameters of the skewed-normal distributions. Since the "E step" in the proposed algorithm has a closed form of weighted linear cominations of the observed absornance, the result can be generalized to data observed simulateously at two or more wave lengths of light using the known absorbance ratios. The method is illustrated with liquid chromatographic data on opioids.