Handling experimental data in FACPACK: noise effects on duality relations and algorithmic parameter selection

Abstract

The area of feasible solutions (AFS) quantifies factor ambiguity in multivariate curve resolution (MCR). Although AFS computation is well-defined for noise-free data, experimental datasets require constraint relaxation and parameter tuning, which influence the resulting feasible bands. This work addresses the computation of the AFS in the presence of noise. We propose a statistically motivated strategy for control parameter estimation based on Savitzky–Golay filtering, introduce a two-stage preprocessing approach for baseline-distorted data, and analyze the effect of noise in the dual space. The methods are demonstrated on model data and experimental FT-IR datasets.