Mean-Structure and Autocorrelation Consistent Covariance Matrix Estimation

A nonparametric estimator for asymptotic covariance matrices in nonstationary time series, robust against unknown trends and divergent change points. The method is algorithmically fast, requiring no change point search, trend estimation, or cross-validation, and includes an automatic optimal bandwidth selector. Empirical studies on four stock market indices demonstrate its statistical and computational efficiency.

Use Cases

• Change point detection in financial time series based on the proposed robust covariance estimator.
• Construction of simultaneous confidence bands for trends in nonstationary data.
• Statistical inference in econometric models where the number of structural breaks is unknown or divergent.
• Benchmarking new time series covariance estimation methods against the proposed nonparametric approach.

Strengths

• Proposed estimator is robust against unknown forms of trends and a potentially divergent number of change points.
• Algorithm is computationally efficient, requiring no search for change points, estimation of trends, or cross-validation.
• Includes an automatic optimal bandwidth selector, enhancing statistical efficiency.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Kin Wai Chan via paperswithcode

Collection Method

Proposed statistical methodology with empirical studies on four stock market indices.