Parameters

Periods to be checked can be passed as parameters, thus eliminating the need for reference data. As an example, an user who only wants to make sure that his monthly data have a seasonal dynamics will only need to specify periods=12. The reference data will be ignored and all actual data periods will be checked against 12 as explained above, with a strength of 1. More periods can be passed by using commas to separate the numbers, without spaces, as in periods=6,12.

If difference=1 is passed, the data are differenced before computing. This is a commonly used method to clean data from obvious trends.

An acceptable drift (in time units) within which periods are considered equivalent (but still weighted according to distance) can be specified with the acceptable_drift=n parameter, whose default value is 1.

The peak detection algorithm uses heuristics to detect the ``significant'' peaks in the periodogram; the default behavior can be changed by means of three parameters. peak_threshold=x sets the relative threshold of non-significance: any peak lower than x will not be considered. The default is 0.1 (all parameters refer to normalized peaks, thus vary between 0 and 1). This only influences the speed of the algorithms and does not need to be changed under normal conditions. Two parameters define the minimum height and maximum width of the peaks compared to the span of the periodogram. The default value for peak_height, the minimum relative height of a peak in the normalized power spectrum, is 0.5; the default of peak_width, the maximum width of a peak compared to the span of the whole periodogram, is 0.1. The length of the periodogram is $1/2$ of the first power of 2 which is greater or equal than the total number of data or reference data, whichever is less. ``Significant'' periods longer than $n/2$ are not considered.

By default, no more than $4$ peaks in the reference data are considered. In some occasions an user might want to check more or less, for example considering only one peak to check the principal frequency in data. If max_peaks=n is passed, the maximum number of peaks considered is set to $n$. The peaks will be ordered according to strength, so the ones used will be the strongest ones.