incline.local_sigma

incline.local_sigma(y, window_length=25)

Estimate a noise level that varies across the series.

Applies the same second-difference logic as rice_sigma() inside a rolling window, so the estimate stays free of the smoother and free of the trend while being allowed to change.

Parameters:

• y (NDArray[float64]) – Observed values.

• window_length (int) – Number of second differences averaged at each point. Wider is steadier but slower to follow a change in scale.

Returns:

Estimated noise standard deviation at each point.

Raises:

ValueError – If y is not one-dimensional or window_length is not an odd integer of at least three.

Return type:

NDArray[float64]