<a id="skfolio-attribution-component"></a>

# skfolio.attribution.Component

<a id="skfolio.attribution.Component"></a>

### *class* skfolio.attribution.Component(vol_contrib, pct_total_variance, mu_contrib, vol, corr_with_ptf, mu_uncertainty=None)

Portfolio attribution component.

Represents one component of the portfolio attribution: systematic, idiosyncratic,
unattributed, or total. Each component stores volatility and return contributions,
the percentage of total portfolio variance, standalone volatility, correlation
with the portfolio and optional return uncertainty.

For single-point attribution, fields are floats. For rolling attribution (from
[`rolling_realized_factor_attribution`](https://skfolio.org/generated/skfolio.attribution.rolling_realized_factor_attribution.html.md#skfolio.attribution.rolling_realized_factor_attribution)), fields are 1D arrays of shape
`(n_windows,)`.

* **Attributes:**
  **vol_contrib** *float or ndarray of shape (n_windows,)*
  : Volatility contribution to total portfolio volatility.

  **pct_total_variance** *float or ndarray of shape (n_windows,)*
  : Percentage of total portfolio variance.

  **mu_contrib** *float or ndarray of shape (n_windows,)*
  : Return contribution to total portfolio return (expected return for predicted
    attribution and mean return for realized attribution).

  **vol** *float or ndarray of shape (n_windows,)*
  : Standalone component volatility.

  **corr_with_ptf** *float or ndarray of shape (n_windows,)*
  : Correlation with portfolio returns.

  **mu_uncertainty** *float or ndarray of shape (n_windows,) or None*
  : Standard error of the mean return attribution, reflecting estimation uncertainty
    in the cross-sectional factor return regression. The systematic and
    idiosyncratic values are equal because their estimation errors sum to zero
    (the total portfolio return is observed). `None` when uncertainty is not
    computed.

<a id="skfolio.attribution.Component.mu"></a>

#### *property* mu

Standalone component return (expected return for predicted attribution and
mean return for realized attribution).

Components do not store a separate standalone return statistic. Unlike `vol`,
the component-level return is already its contribution to total portfolio
return, so `mu` is equal to `mu_contrib`.