Effect of sample complexity on quantification of analytes in aqueous samples by near-infrared spectroscopy

This study was undertaken to quantitate the impact of increasing sample com plexity on near infrared spectroscopic (NIRS) measurements of small molecul es in aqueous solutions with varying numbers of components. Samples with 2, 6, or 10 varying components were investigated. Within the 10-component sam ples, three analytes were quantified with errors below 6% and seven of the analytes were quantified with errors below 10%. An increase in the number o f varying components can substantially increase the error associated with m easurement, A comparison of measurement errors across sample sets, as gauge d by the standard error of prediction (SEP), reveals that an increase in th e number of varying components from 2 to 6 increases the SEP by approximate ly 50%. An increase from 2 to 10 varying components increases the SEP by ap proximately 340%. While there appear to be no substantial correlations betw een the presence of a specific analyte and the errors associated with quant ification of another analyte, several analytes do display a small degree of sensitivity to varying concentrations of certain background components. Th e analysis also demonstrates that calibrations containing an overestimation of the numbers of varying components can substantially increase measuremen t errors and so calibrations must be constructed with an accurate understan ding of the number of varying components that are likely to be encountered.