Circadian Temperature Rhythm Power Spectra: Is Equal Sampling Necessary?

This paper uses the rectal temperature data from 7 free‐running experiments (132 subject‐days), and 3 forced desynchronization experiments (109 subject‐days) to address the question of whether equal spacing of data points is truly needed in order to estimate the “power spectra’ of circadian rhythmicity in that measure, and thus tau, the best fitting period. In particular, it compares a standard periodogram analysis of minute‐by‐minute temperatures, with a cosinor‐based analysis using only about six readings per day, collected at irregular intervals, limited to the time of wakefulness of the subject. In each case “power spectra’ were created for periods between 22.0h and 27.0h in steps of 0.1h. The results suggested that if there is a sufficiently large number of cycles (>14), and if interest is restricted to the “power spectrum’ and/or tau, then equal spacing is probably not required. In 9 out of the 10 subjects, the two techniques produced almost identical “power spectra’ and estimates of best‐fitting period, with the correlation coefficient between the two “power spectra’ exceeding .8, and the difference between the two estimates of tau never exceeding 18 min.