Evaluation of results obtained by analysis of fractions

The paper discusses treatment of experimental data obtained from the monitoring of continuous processes by a series of batch analyses (i.e. by analysis of fractions sequentially collected during the process). Common errors that are committed when graphically representing and fitting such data are shown. An approach for identification of such errors in published data is presented. One error results from the incorrect positioning of the analysis results along the X-axis when plotting data. In the case of equal fractions, this results in a systematic error along the X-axis. For non-equal fractions, the error in the coefficients of the fitting equation can be both systematic and random. A correct representation does not introduce any error if the actual function is linear in the region of the fraction collection. This plotting method is also a satisfactory procedure for low curvature functions. Another error originates because of the incorrect application of the least squares method. The reason lies in the fact that the analysis of each fraction does not represent an instant value of the measured parameter, but rather the averaged value over some period of time or over some volume. The actual error depends on the curvature of the fitting function. Furthermore, a reader encounters difficulties when evaluating published scientific information because many authors do not explain the methods by which their data were plotted and fit. The presented approach to the evaluation of graphical data makes it possible to identify the described errors.