connectedness 0.1.0a0

Diebold-Yilmaz connectedness and volatility spillover indices in Python.

connectedness

Diebold-Yilmaz connectedness and volatility spillover indices in Python.

connectedness turns a panel of asset prices into a fully reproducible spillover analysis: realized covariance and semicovariance matrices, optional PIT normality transform, MHAR-LASSO dynamics, generalized FEVD (Koop-Pesaran-Shin), and the static and rolling-window spillover indices of Diebold and Yilmaz.

Status

Alpha. The API may change before 1.0.0.

Installation

pip install connectedness            # core
pip install connectedness[plots]     # + matplotlib helpers
pip install connectedness[parallel]  # + joblib/tqdm for rolling windows
pip install connectedness[io]        # + pyarrow for parquet I/O

Quickstart

import pandas as pd
from connectedness import SpilloverPipeline

# wide DataFrame: DatetimeIndex (intraday) × markets
prices = pd.read_parquet("prices.parquet")

result = (
    SpilloverPipeline(prices)
    .compute_returns()
    .compute_realized_measures()
    .auto_pit()
    .fit_static()
    .fit_rolling(window=365)
    .run()
)

print(result.tsi)              # total spillover index (%)
print(result.directional_to)   # contributions TO others
print(result.directional_from) # contributions FROM others
print(result.rolling_tsi)      # rolling TSI time series

For the underlying primitives:

from connectedness import (
    simple_returns,
    realized_covariance,
    realized_semicov,
    pit_transform,
    static_spillover,
    rolling_spillover,
)

Methodology

See docs/methodology.md for the formal definitions of realized (semi)covariance, the MHAR-LASSO specification, the generalized variance decomposition, and the Diebold-Yilmaz total / directional / net spillover indices.

Citation

@software{connectedness,
  author  = {Rolotti, Francisco},
  title   = {connectedness: Diebold-Yilmaz spillover indices in Python},
  year    = {2026},
  url     = {https://github.com/franrolotti/connectedness}
}

See also CITATION.cff.

License

MIT. See LICENSE.