Biological time-series analysis via a fast orthogonal search

Advantages of time-series analysis are described and illustrated by the fast orthogonal search method. The method can be used to approximate biological time-series data by a parsimonious sinusoidal series model or representation. The component frequencies in this model need not be commensurate nor integral multiples of the fundamental frequency corresponding to the record length. The method achieves economy of representation by finding the most significant frequencies first. In simulations, the author shows that the method really copes with missing or unequally-spaced data, and is capable of five to eight time the frequency resolution of a conventional Fourier-series analysis.<
>