Covariance Matrices

Covariance matrices are used in FIPS to represent uncertainty in the data provided to the inverse problem.

CovarianceMatrix

Constructor

CovarianceMatrix(data[, name, index, ...])

Represents a symmetric Covariance Matrix.

Attributes and underlying data

CovarianceMatrix.name	Name of the structure, used for error messages and block naming.
CovarianceMatrix.data	The underlying data, which must be numeric and non-NaN.
CovarianceMatrix.index	Return the index of the underlying data.
CovarianceMatrix.columns	Return the columns of the underlying DataFrame.
CovarianceMatrix.shape	Return the shape of the underlying data.
CovarianceMatrix.values	Return values as numpy array or sparse matrix.
CovarianceMatrix.variances	Returns the variances (diagonal elements) of the covariance matrix.
CovarianceMatrix.blocks	Accessor for retrieving MatrixBlock instances from the Matrix.

Conversion

CovarianceMatrix.copy([deep])	Create a copy of the structure.
CovarianceMatrix.to_frame()	Get the underlying DataFrame data.
CovarianceMatrix.to_numpy()	Get the underlying data as a NumPy array.

Reindexing / selection / label manipulation

CovarianceMatrix.xs(key[, axis, level, ...])	Cross-select data based on index/column values.
CovarianceMatrix.reindex(index[, columns, ...])	Return a new instance with data reindexed to the specified index and columns.
CovarianceMatrix.round_index(decimals[, ...])	Round float indices on the specified axis to given decimals.

Serialization

CovarianceMatrix.from_file(path)	Load object from a pickle file.
CovarianceMatrix.to_file(path)	Save object to a pickle file.

Sparse support

CovarianceMatrix.is_sparse	True if the internal DataFrame uses pandas sparse storage.
CovarianceMatrix.to_sparse([threshold])	Return a copy with sparse internal storage.
CovarianceMatrix.to_dense()	Return a copy with dense internal storage.

Computations

CovarianceMatrix.force_symmetry([keep])

Force the matrix to be perfectly symmetric by copying one triangle to the other.

ErrorComponents

ErrorComponent(name, variances)	Abstract base class for error components in covariance matrices.
DiagonalError(name, variances)	Error component that produces diagonal covariance matrices.
BlockDecayError(name, variances, groupers, ...)	Error component with block-structured correlation decay.
KroneckerError(name, variances, marginal_kernels)	Builds a full covariance matrix for a strict grid.

Build covariance matrix

ErrorComponent.build(index, **kwargs)

Must return a pd.DataFrame with the given index on both rows and columns.

Binary operations

ErrorComponent.__add__(other)

Add error components to create a CovarianceBuilder.

CovarianceBuilder

Constructor

CovarianceBuilder(components)

Builds covariance matrices by combining multiple error components.

Build covariance matrix

CovarianceBuilder.build(index[, sparse])

Build and sum all error components into a single DataFrame.

Binary operations

CovarianceBuilder.__add__(other)	Add error components or builders together.
CovarianceBuilder.__radd__(other)	Right-hand addition for error components.