Extracting features from experimental data.

Univariate polynomial spline curves provide a flexible class of functions that are effective for modelling a wide variety of experimental data. However, the parameters defining such curves generally do not provide directly any physical information about the measurement system giving rise to the data. Instead such information is required to be extracted from the fitted model. The problem of extracting information from univariate polynomial spline curves is considered, where that information takes the form of features of the curve, including the positions of zero-crossing points, peaks, troughs and points of inflexion, and the width of peaks and troughs. The evaluation of the uncertainties associated with estimates of these features derived from a spline curve fitted to experimental data is addressed.