Laboratory instrumentation and computing: comparison of six methods for the determination of the period of circadian rhythms.

Data sets with known periodicity were used to compare the accuracy and noise tolerance of six methods of circadian period analysis: Fourier analysis, autocorrelation, Enright's (chi-square) periodogram, linear regression of onsets, interonset averaging, and acrophase counting. All methods except acrophase counting accurately detected the period of rhythms with pure waveform (cosine and square wave), whereas Fourier analysis and Enright's periodogram were superior to the other methods in the analysis of more complex waveforms (which more closely resemble actual circadian rhythms). The sensitivity of all methods was reduced by the insertion of random noise into the original data sets, but the methods of autocorrelation and Enright's periodogram were more tolerant of low signal-to-noise ratios than the remaining methods. Although particular situations may require particular methods, the results suggest that Enright's periodogram is the best choice as a general method for the determination of the period of circadian rhythms.