Systematic comparison of data-processing options for kinetic-based, single-component determinations of noncatalysts: Effects of random noise

This paper describes effects of random noise on results obtained by applying a variety of kinetic-based data-processing options to simulated first-order data for a variety of noise amplitudes, starting points, data ranges and number of data points. The first part of the study focuses on a curve-fitting predictive option and subsequent parts focus on comparisons among several data-processing options. Whenever possible, empirical equations are used to quantify effects of the several variables. For several multipoint options, it was found that a single equation can be used to describe combined effects of noise amplitude, number of data points and fitting range on the imprecision of computed signal change. An analogous equation can be used to describe the effect of the starting point of the curve-fitting process on the imprecision of signal changes computed with the predictive option. No such equation was found for combined effects of these parameters on the inaccuracy of computed signal changes. The final phase of the study focused on effects of signal noise on simulated calibration results for a variety of data-processing options. For most situations studied, it is found that the imprecision and inaccuracy of results are virtually independent of signal amplitude but increase linearly with the noise amplitude. Sensitivities of the imprecision and scatter of points on calibration plots to noise amplitude were lowest for a two-point/fixed-time option; slightly larger (~2-fold) for one-rate, partial-sums, successive-integration and predictive options; and much larger (≥ 10-fold) for three- and four-point/fixed-time and two-rate options. Noise sensitivities of the Kezdy-Swinbourne and Guggenheim options are similar to the one-rate, partial-sums, successive-integration and predictive options for small noise and large signal amplitudes but are similar to the three- and four-point fixed-time and two-rate options for larger noise and smaller signal amplitudes.