ExpandingSkew

Description

The ExpandingSkew function returns the adjusted Fisher-Pearson standardized moment coefficient (G1) of every sample seen since the last reset. It uses the same bias-corrected estimator as RollingSkew and pandas.Series.expanding().skew(). Undefined (NaN) until at least three samples have been seen. Memory is O(1).

Equation:

\[ y[t] = \frac{n}{(n-1)(n-2)}\sum_{i=0}^{t}\left(\frac{x[i]-\bar{x}}{s}\right)^3, \quad n=t+1 \]

Parameters: none.

NaN handling

Policy: ignore. A NaN in the input at index t causes the function to skip that step: output at t is NaN and internal state is unchanged. Subsequent finite samples are processed as if step t had not occurred.

Examples

Usage example

import numpy as np
from screamer import ExpandingSkew

x = np.arange(1.0, 11.0)
y = ExpandingSkew()(x)