incline.Heteroskedastic¶
- class incline.Heteroskedastic(sigma=None, window=25)[source]¶
Independent noise whose scale changes across the series.
Assuming one constant noise level when the scale varies gets the average right and everything else wrong, in opposite directions at the two ends. Measured on a series whose noise standard deviation rises from 0.2 to 1.0, with a nominal 95% interval: coverage 1.000 and a standard error 2.0x too large in the quiet stretch, coverage 0.828 and a standard error 0.70x too small in the noisy one. The midpoint looks perfect, which is what makes it easy to miss.
- Variables:
sigma (numpy._typing._array_like.NDArray[numpy.float64] | None) – Per-point standard deviations. Estimated locally when None.
window (int) – Points averaged by the local estimator.
- Parameters:
sigma (NDArray[float64] | None)
window (int)
- __init__(sigma=None, window=25)¶
- Parameters:
sigma (NDArray[float64] | None)
window (int)
- Return type:
None
Methods
__init__([sigma, window])estimate(y, axis)Estimate the local noise level, or use the supplied one.
Attributes