Effects of data analysis on accuracy and precision of GPC results

Abstract

AbstractThe effects of data imprecision and baseline uncertainties have been investigated by computer simulation of GPC data from a polymer sample with a truncated log‐normal molecular weight distribution. If the data are very accurate, as few as five data points can be sampled without serious error in computed molecular weight averages. The number of data points required is much larger, however, if these are taken at equal increments of molecular weight rather than elution volume. The effects of noise can be counteracted by increasing the sampling frequency. Baseline uncertainties present a more serious problem, especially for broad‐distribution samples. If the detected signal is too noisy to permit accurate location of the baseline, errors can be minimized by using a second, more sensitive detector to determine the peak start and end. It is very difficult to estimate M̄z and higher molecular weight averages accurately if the noise level is greater than 0.5%.