Point perturbation analysis of experimental data

A new method of data analysis is proposed. The method is based on discrete perturbation of experimental data points, which is used to probe the metric of the parameter hyperspace. Perturbation-induced fluctuations in the residual values are analysed by discrete Fourier transform to yield the autocorrelation function and a relaxation length for each experimental point. This parameter provides a quantitative measure of correlation and hence nonrandomness of residuals. The method is applied to the analysis of measurements of the shear viscosity of a 2,6-lutidine/water mixture near the critical point, and to the oxygen and carbon monoxide binding reactions to human hemoglobin. Relaxation profiles are constructed for several experimental data sets. Departure from random behavior in the residuals is discussed in connection with the theoretical interpretations of the phenomenon under consideration.