(285a) Validation of Noisy Single-Exponential Decay Data Using a Residual Bootstrap and Moving Average Subtraction (VRBMAS) Technique
The presence of noise often masks the presence of nonrandom deviations in experimental data and thus falsify fitting and validation efforts. An omnibus validation technique using residual bootstrap and moving average subtraction (VRBMAS) is developed to analyze noisy, single-exponential decay data; by decoupling the nonrandom deviations from the random noise in the residuals, departures from exponentiality can be detected. VRBMAS is compared to two existing methods, the Chi-squared and the runs tests. All three methods seem to be unbiased over a limited range of levels of significance. VRBMAS seems to be the most powerful validation technique and is more sensitive to longer period irregularities in the residual plot while the runs test is better suited for detecting high frequency correlations.
The initial part of this work was carried out at Rensselaer Polytechnic Institute (RPI) and supported by the U.S. Army under Grants DAAG55-98-1-0183 and DAAD 19-99-1-0204. Portions of this work were performed at Argonne National Laboratory under the auspices of the Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences, and Biosciences, U.S. Department of Energy, under contract number DE-AC02-06CH11357.