pyhctsa.operations.stationarity.moment_corr¶

pyhctsa.operations.stationarity.moment_corr(x, window_length=None, w_overlap=None, mom_1='mean', mom_2='std', what_transform='none')¶

Correlations between simple statistics in local windows of a time series. The idea to implement this was that of Prof. Nick S. Jones (Imperial College London).

Parameters:¶

x : array-like¶

The input time series.

window_length : float, optional¶

The sliding window length (can be a fraction to specify or a proportion of the time-series length). Default is None.

w_overlap : None | float¶

The overlap between consecutive windows as a fraction of the window length. Default is None.

mom_1 : str, optional¶

The statistics to investigate correlations between (in each window)

• ’iqr’: interquartile range

• ’median’: median

• ’std’: standard deviation (about the local mean)

• ’mean’: mean

Default is 'mean'.

mom_2 : str, optional¶

The statistics to investigate correlations between (in each window)

• ’iqr’: interquartile range

• ’median’: median

• ’std’: standard deviation (about the local mean)

• ’mean’: mean

Default is 'mean'.

what_transform : str, optional¶

The pre-processing what_transform to apply to the time series before analyzing it

• ’abs’: takes absolute values of all data points

• ’sqrt’: takes the square root of absolute values of all data points

• ’sq’: takes the square of every data point

• ’none’: does no what_transform

Default is 'none'.

Returns:¶

Dictionary of statistics related to the correlation between simple statistics in local windows of the input time series: the correlation coefficient R, its absolute value absR, and density, the number of windows divided by the bounding-box area in the M1–M2 plane.

Return type:¶

out

Notes

Only full windows are used: windows start every window_length - w_overlap samples from the first sample, and a final window that would run past the end of the series is not included.