add test for gaussian

Author: lionelkusch

src/hidimstat/statistical_tools/gaussian_distribution.py

@@ -44,12 +44,7 @@ def __init__(self, cov_estimator, random_state=None, centered=False, tol=1e-14):
         self.cov_estimator = cov_estimator
         self.centered = centered
         self.tol = tol
-        if isinstance(random_state, (int, np.int32, np.int64)):
-            self.rng = check_random_state(random_state)
-        elif random_state is None:
-            self.rng = check_random_state(0)
-        else:
-            raise TypeError("Wrong type for random_state")
+        self.rng = check_random_state(random_state)
 
     def fit(self, X):
         """

test/statistical_tools/test_gaussian_distribution.py

@@ -0,0 +1,85 @@
+import numpy as np
+import pytest
+from hidimstat.statistical_tools.gaussian_distribution import (
+    _s_equi,
+    GaussianDistribution,
+)
+from hidimstat._utils.scenario import multivariate_simulation
+from sklearn.covariance import LedoitWolf
+
+
+def test_gaussian_equi():
+    """test function of gaussian"""
+    seed = 42
+    n = 100
+    p = 50
+    X, y, beta, noise = multivariate_simulation(n, p, seed=seed)
+    generator = GaussianDistribution(
+        cov_estimator=LedoitWolf(assume_centered=True),
+        random_state=seed * 2,
+        centered=False,
+    )
+    generator.fit(X=X)
+    X_tilde = generator.sample()
+    assert X_tilde.shape == (n, p)
+
+
+def test_gaussian_center():
+    """test function of gaussian"""
+    seed = 42
+    n = 100
+    p = 50
+    X, y, beta, noise = multivariate_simulation(n, p, seed=seed)
+    generator = GaussianDistribution(
+        cov_estimator=LedoitWolf(assume_centered=True),
+        random_state=seed * 2,
+        centered=True,
+    )
+    generator.fit(X=X)
+    X_tilde = generator.sample()
+    assert X_tilde.shape == (n, p)
+
+
+def test_gaussian_error():
+    """test function error"""
+    seed = 42
+    n = 100
+    p = 50
+    X, y, beta, noise = multivariate_simulation(n, p, seed=seed)
+    generator = GaussianDistribution(
+        cov_estimator=LedoitWolf(assume_centered=True),
+        random_state=seed * 2,
+        centered=True,
+    )
+    with pytest.raises(
+        ValueError, match="The GaussianGenerator requires to be fit before simulate"
+    ):
+        generator.sample()
+
+
+def test_s_equi_not_define_positive():
+    """test the warning and error of s_equi function"""
+    n = 10
+    tol = 1e-7
+    seed = 42
+
+    # random positive matrix
+    rgn = np.random.RandomState(seed)
+    a = rgn.randn(n, n)
+    a -= np.min(a)
+    with pytest.raises(
+        Exception, match="The covariance matrix is not positive-definite."
+    ):
+        _s_equi(a)
+
+    # matrix with positive eigenvalues, positive diagonal
+    while not np.all(np.linalg.eigvalsh(a) > tol):
+        a += 0.1 * np.eye(n)
+    with pytest.warns(UserWarning, match="The equi-correlated matrix"):
+        _s_equi(a)
+
+    # positive definite matrix
+    u, s, vh = np.linalg.svd(a)
+    d = np.eye(n)
+    sigma = u * d * u.T
+    _s_equi(sigma)