Corrections for heteroscedasticity in window evolving factor analysis

Window evolving factor analysis (WEFA) is a powerful technique for checking peak purity in liquid chromatography with diode array detection (LC-DAD). However, practical application of the technique can be limited by instrumental and experimental non-idealities. One of the problem sources is heteroscedasticity. In this work, we propose two new data transformation procedures and one technique for directly adjusting the log-eigenvalue profiles, which eliminate most of the heteroscedastic effect in the WEFA plots. The pretreatment and the adjustment techniques can be used in combination to obtain even better results. The performance of the techniques are demonstrated on simulated and actual data.