This code is capable of calculating the difference between two compressed covariance matrices. It follows the procedure described in Tassia & Valerio 2020.
The compressed matrices were compressed using MOPED.
How to use:
import numpy as np
from GetDiff import FindDiff
# Our covariances are in a .txt file, so we can import them as:
CompCovariance = np.loadtxt('CompressedCovariance.txt', dtype=np.float64)
CompGaussianCovariance = np.loadtxt('CompressedGaussianCovariance.txt', dtype=np.float64)
blah = FindDiff(CompGaussianCovariance, CompCovariance, sample_size=500)
# Choose a small step size for testing, but if you want proper results, then
# make sure this is > 1000. It will take longer, so be patient.
# Get the difference with the associated error (68%).
blah.diff
# Returns dictionary where:
# blah.diff['diag'] is the difference for the diagonal elements.
# blah.diff['corr] is the difference for the correlation matrix.
blah.chain
# Returns dictionary with a 2D numpy array with the first column corresponding to
# the chi2 value, and the second column to the respective value of the difference.
# blah.chain['diag'] for the diagonal elements.
# blah.chain['corr] for the correlation matrix.The plots are WIP.
Feel free to use, but please cite Tassia & Valerio 2020. If you have comments, questions, or feedback, please contact Tassia (tassia.aferreira@gmail.com)