The Number of Components in Noisy Data from a Swept-Potential Electrochemical Detector

Experimental data obtained from a conventional HPLC using a swept-potential electrochemical detector are analyzed for the number of components present under varying conditions of signal-to-noise (S/N) ratios. The autocorrelation, multiple determination, and single-vector uniqueness test functions are evaluated under low S/N conditions, and also tested on Fourier-smoothed data. The autocorrelation and multiple determination functions give good predictions down to the S/N ratio of 2.4. Below this value the autocorrelation function gives fewer counts of the components than expected as the components signal becomes buried in the background noise. On the other hand, the multiple determination function gives too many counts of the components actually present. The single-vector uniqueness test function proves to be superior to the other two. Its predictions are more consistent even at S/N ratios lower than 2.4 Data smoothing significantly improves the performance of all three functions