roll_risk_contribution¶

Defined in fynance.portfolio.attribution

roll_risk_contribution(W, X, n=252, cov=None, pct=True)[source]

Causal rolling risk contributions over a time series.

For each time step \(t \geq n\), estimate the covariance matrix from returns in the past \(n\) periods and compute risk contributions for the weights at time \(t\). Earlier rows are filled with NaN (no covariance estimate yet).

Parameters:

Warray_like: Weight time series, shape (T, N).
Xarray_like: Returns panel, shape (T, N), rows in chronological order (oldest first).
nint, optional: Window length for covariance estimation. Default 252.
covcallable, optional: Covariance estimator callable that accepts a returns panel and returns a symmetric (N, N) matrix. If None (default), uses numpy.cov with rowvar=False.
pctbool, optional: If True (default), return percentage contributions summing to 1. If False, return absolute contributions summing to portfolio volatility per period.

Returns:

np.ndarray: Risk contributions per period, shape (T, N). Rows t < n are filled with NaN.

Examples

>>> import numpy as np
>>> rng = np.random.default_rng(0)
>>> X = rng.standard_normal((50, 3))
>>> W = rng.uniform(0.2, 0.4, (50, 3))
>>> W /= W.sum(axis=1, keepdims=True)
>>> rc = roll_risk_contribution(W, X, n=20)
>>> rc.shape
(50, 3)
>>> np.isnan(rc[:20]).all()
True