pycorrelation 0.0.2

PyCorrelation is a Python package designed to handle symmetric and correlation matrices efficiently.

PyCorrelation

PyCorrelation is a Python package designed to handle symmetric and correlation matrices efficiently. It builds upon the concept of symmetric matrices to provide a class dedicated to correlation operations, ensuring values fall within the valid range $(-1,1)$, and automatically maintaining correlations of 1 on the diagonal.

Features

• A class ensuring symmetry in a matrix: $x_{i,j} = x_{j,i}$
• A class for managing correlation matrices with strict validation of correlation values.
• Methods to set, retrieve, and validate correlation pairs using user-friendly indexers.

Installation

To install PyCorrelation, you can clone the repository and use pip:

git clone https://github.com/SRKX/PyCorrelation.git
cd PyCorrelation
pip install .

Usage

Below is an example of how to use the package:

from PyCorrelation import CorrelationMatrix

# Initialize a CorrelationMatrix
correl_matrix = CorrelationMatrix(keys=['A', 'B', 'C'], frozen_keys=True)

# Set correlations
correl_matrix[('A', 'B')] = 0.5
correl_matrix[('B', 'C')] = -0.2

# Access correlation values
print(correl_matrix[('B', 'A')])  # Outputs: 0.5

# Check if a key pair is in the matrix
print(('A', 'B') in correl_matrix)  # Outputs: True

Classes and Methods

SymmetricMatrix

This is the base class for symmetric matrices. It allows you to define and manipulate a matrix where elements are mirrored across the diagonal.

Public Methods

• keys(): Returns the keys of the matrix.
• to_2d_dict(selected_keys=None): Converts the matrix into a 2D dictionary, extracting only the selected keys (or all if not specified).
• to_2d_array(ordered_keys): Converts the matrix into a 2D array representation, returning a list of lists where the indices follow the ordered of the input parameter.

CorrelationMatrix

This class extends SymmetricMatrix and is specialized for correlation matrices. It includes additional validation to ensure values are always in $(-1,1)$, and that identical keys always have a correlation of 1.

License

This project is licensed under the MIT License.