Comparison of Covariance Estimators

Introduction

Covariance estimation is an important task in statistical analysis. In this lab, we will compare two methods of covariance estimation: Ledoit-Wolf and OAS. We will use Gaussian distributed data to compare the estimated MSE of these two methods.

VM Tips

After the VM startup is done, click the top left corner to switch to the Notebook tab to access Jupyter Notebook for practice.

Sometimes, you may need to wait a few seconds for Jupyter Notebook to finish loading. The validation of operations cannot be automated because of limitations in Jupyter Notebook.

If you face issues during learning, feel free to ask Labby. Provide feedback after the session, and we will promptly resolve the problem for you.

Skills Graph

Import Libraries

First, we need to import the necessary libraries for this lab. We will be using numpy for numerical calculations, matplotlib for visualizations, and scikit-learn for covariance estimation.

import numpy as np
import matplotlib.pyplot as plt
from scipy.linalg import toeplitz, cholesky
from sklearn.covariance import LedoitWolf, OAS

Generate Data

Next, we will generate Gaussian distributed data with a covariance matrix that follows an AR(1) process. We will use toeplitz and cholesky functions from scipy.linalg to generate the covariance matrix.

np.random.seed(0)

n_features = 100
r = 0.1
real_cov = toeplitz(r ** np.arange(n_features))
coloring_matrix = cholesky(real_cov)

Compute MSE and Shrinkage

We will compare the Ledoit-Wolf and OAS methods using the simulated data. We will compute the mean squared error (MSE) and shrinkage of both methods.

n_samples_range = np.arange(6, 31, 1)
repeat = 100
lw_mse = np.zeros((n_samples_range.size, repeat))
oa_mse = np.zeros((n_samples_range.size, repeat))
lw_shrinkage = np.zeros((n_samples_range.size, repeat))
oa_shrinkage = np.zeros((n_samples_range.size, repeat))

for i, n_samples in enumerate(n_samples_range):
    for j in range(repeat):
        X = np.dot(np.random.normal(size=(n_samples, n_features)), coloring_matrix.T)

        lw = LedoitWolf(store_precision=False, assume_centered=True)
        lw.fit(X)
        lw_mse[i, j] = lw.error_norm(real_cov, scaling=False)
        lw_shrinkage[i, j] = lw.shrinkage_

        oa = OAS(store_precision=False, assume_centered=True)
        oa.fit(X)
        oa_mse[i, j] = oa.error_norm(real_cov, scaling=False)
        oa_shrinkage[i, j] = oa.shrinkage_

Plot Results

Finally, we will plot the results to compare the MSE and shrinkage of the Ledoit-Wolf and OAS methods.

plt.subplot(2, 1, 1)
plt.errorbar(
    n_samples_range,
    lw_mse.mean(1),
    yerr=lw_mse.std(1),
    label="Ledoit-Wolf",
    color="navy",
    lw=2,
)
plt.errorbar(
    n_samples_range,
    oa_mse.mean(1),
    yerr=oa_mse.std(1),
    label="OAS",
    color="darkorange",
    lw=2,
)
plt.ylabel("Squared error")
plt.legend(loc="upper right")
plt.title("Comparison of covariance estimators")
plt.xlim(5, 31)

plt.subplot(2, 1, 2)
plt.errorbar(
    n_samples_range,
    lw_shrinkage.mean(1),
    yerr=lw_shrinkage.std(1),
    label="Ledoit-Wolf",
    color="navy",
    lw=2,
)
plt.errorbar(
    n_samples_range,
    oa_shrinkage.mean(1),
    yerr=oa_shrinkage.std(1),
    label="OAS",
    color="darkorange",
    lw=2,
)
plt.xlabel("n_samples")
plt.ylabel("Shrinkage")
plt.legend(loc="lower right")
plt.ylim(plt.ylim()[0], 1.0 + (plt.ylim()[1] - plt.ylim()[0]) / 10.0)
plt.xlim(5, 31)

plt.show()

Summary

In this lab, we compared the Ledoit-Wolf and OAS methods for covariance estimation using Gaussian distributed data. We plotted the MSE and shrinkage of both methods and found that the OAS method has better convergence under the assumption that the data are Gaussian.