Rectangular harmonic analysis revisited

Abstract

Alldredge's method of rectangular harmonic analysis has been reexamined. After correction of errors, it is found to give improbable values between the data points and wild values outside them. A much more realistic model has been obtained by (1) determining only the most significant coefficients.(those that exceed their standard deviations, obtained by an iterative process), (2) introducing new parameters to allow for a linear trend across the region, and (3) increasing the scaling factors so that the sinusoids start and finish outside the region. Modification 1 is the most important for improving the interpolative qualities of the model. Modifications 2 and 3 reduce, but do not entirely eliminate, the wild values near the edges.