pyhctsa.utils.detrend_linear¶

pyhctsa.utils.detrend_linear(x)¶

Remove the least-squares line from a time series.

Parameters:¶

x : array-like¶

The time series (1-D).

Returns:¶

float64 array of the same length: x minus the ordinary-least-squares line fitted against the sample index, so the result has zero mean and no linear trend. A series of fewer than 3 samples is returned as zeros (a line passes through any two points).

Return type:¶

numpy.ndarray

Notes

Only the mean and the slope are removed: long memory, slow oscillations and level shifts stay in the series. This is the transform behind the *_lindt features (a feature configuration with preprocess: detrend, which re-z-scores the residual); apply it to every series of a dataset to detrend the data before feature extraction.