A package for estimating and regularising correlation and covariance matrices with high frequency financial data
Author s-baumann
12 Stars
Updated Last
1 Year Ago
Started In
December 2020


Build Coverage Documentation
Build status codecov docs-latest-img

This package offers a number of algorithms for the calculations of volatilities, correlation matrices and covariance matrices. Covariances can be calculated using the simple method, a multivariate kernel method, a spectral covariance method, preaveraging of returns and finally by exploiting differences in the volatility of different composition series (generated by adding two observed series).

A number of regularisation algorithms are also implemented including eigenvalue cleaning, mixing with the identity matrix, mapping to the nearest PSD (Positive Semi Definite) matrix and mapping to the nearest valid covariance matrix.

For a paper describing the capabilities of this package see this paper in the Journal of Statistical Software. For a JuliaCon 2021 talk giving an overview of the package click here

Since this paper I have added the CovarianceModel struct and associated functions. Note that these are all experimental and may be changed in future releases. The CovarianceMatrix struct and associated functions (that are written about in the JSS paper) are not likely to be changed however.

Used By Packages

No packages found.