A spectral-analysis program with graphics for the CBM/PET/C-64 microcomputers

Biological rhythms are an important aspect of behavior (Rusak & Zucker, 1975; Saunders, 1977). Circadian rhythms, which have a periodicity of approximately 24 h, have received most attention, but shortand long-term rhythms also playa role in behavior (Broom, 1979, 1980; Campbell & Shipp, 1974, 1979; Daan & Slopsema, 1978). Various methods for detecting such rhythms have been described (Broom, 1979; Campbell & Schipp, 1974; Linkens, 1979; Westman, 1977). Spectral analysis is a technique that is particularly useful for this purpose, and examples of its application are those of Broom (1979); Campbell and Shipp (1974). A spectral-analysis tutorial was provided by Sturgis (1983). The program described here can be used for the analysis ofdata collected by an activity monitor wired to a datalogging system such as those described by Campbell and Forest (1984) and Martin and Unwin (1980) or of any body ofdata collected at equal time intervals (unit time intervals). Long records are desirable (say, > 100), particularly if the record is noisy. The program has been written to operate on the CBM/PET/C-64 microcomputers with all versions of BASIC in conjunction with the various Commodore peripherals. Input can be via keyboard, tape, or disk. Output can be to the screen or printer and can be saved on tape or disk. The graphics display and input/output routines are essentially the same as those used in the data editor and high-resolution histogram plotter described by Campbell (1983). The histogram routine utilizes the graphics set available on the CBM/PET/C-64 microcomputers. An application of spectral analysis was given by Campbell and Shipp (1974); only brief details of the procedure follow. The first step is the removal of longterm trend from the sample values, which in this case is achieved by subtraction of a polynomial regression line from the data to give a series of difference values. For a sample of several hundred readings, a firstor seconddegree polynomial would normally be sufficient for subtraction, but if high spectral density persists at zero frequency, a higher order polynomial could be used. The nth-order regression and F-distribution routines used in this program to obtain the difference values are from Poole and Borchers (1979).