sf_quant.data.construct_covariance_matrix

sf_quant.data.construct_covariance_matrix(date_: date, barrids: list[str]) → DataFrame

Construct the asset covariance matrix from a factor model.

This function builds a covariance matrix for the given assets by combining factor exposures, factor covariances, and specific (idiosyncratic) risks. The resulting covariance matrix can be used as an input to mean-variance optimization.

Parameters

date_datetime.date

The date for which the covariance matrix is computed.

barridslist of str

List of Barrid identifiers for the assets.

Returns

pl.DataFrame

A square covariance matrix stored in a Polars DataFrame.

• Rows and columns are indexed by barrid.

• Column barrid lists the asset identifiers.

• Each subsequent column corresponds to the covariance of the row asset with the column asset.

Notes

The input factor covariance matrix is assumed to be positive semidefinite (PSD).

Examples

>>> import sf_quant.data as sfd
>>> import datetime as dt
>>> date_ = dt.date(2024, 1, 3)
>>> barrids = ['USA06Z1', 'USA0771']
>>> covariance_matrix = sfd.construct_covariance_matrix(
...     date_=date_,
...     barrids=barrids
... )
>>> covariance_matrix
shape: (2, 3)
┌─────────┬─────────────┬──────────────┐
│ barrid  ┆ USA06Z1     ┆ USA0771      │
│ ---     ┆ ---         ┆ ---          │
│ str     ┆ f64         ┆ f64          │
╞═════════╪═════════════╪══════════════╡
│ USA06Z1 ┆ 3224.338938 ┆ 697.641425   │
│ USA0771 ┆ 697.641425  ┆ 11158.366868 │
└─────────┴─────────────┴──────────────┘