fips.covariance.KroneckerError#

class fips.covariance.KroneckerError(name, variances, marginal_kernels)[source]#

Builds a full covariance matrix for a strict grid.

Uses Kronecker products of N arbitrary marginal correlation matrices.

Parameters:

name (str) – Name of the error component.
variances (float or pd.Series) – Variances for the diagonal of the covariance matrix. Can be a single float or a Series indexed by the same MultiIndex as the covariance matrix.
marginal_kernels (list of tuples) –
Each tuple contains: 1. The dimension name(s) as a string or list of strings. 2. The Callable kernel function that takes a DataFrame of those

unique coordinates and returns a 2D correlation matrix.

Note

ORDER MATTERS: These must be provided in the exact order that the dimensions appear in the matrix’s MultiIndex!

Attributes

`marginal_kernels`	List of marginal kernels, each defined by a tuple of dimension name(s) and a kernel function.
`name`	Name of the error component.
`variances`	Variances for the diagonal of the covariance matrix.

Methods

build(index, **kwargs)

Build covariance matrix using product of marginal kernels.