def test_chi2lnlike():
"""
Test the ability of ``orbitize.lnlike.chi2_lnlike()``
to work properly on arrays.
"""
# test with a single model
model = np.zeros((3, 2))
jitter = np.zeros((3, 2))
data = np.ones((3, 2))
errors = np.ones((3, 2))
seppa_indices = [np.array([1])]
chi2 = lnlike.chi2_lnlike(data, errors, None, model, jitter, seppa_indices)
assert chi2.shape == (3, 2)
assert (chi2 == -0.5 * np.ones(
(3, 2)) - np.log(np.sqrt(2 * np.pi * np.ones((3, 2))))).all()
# test with multiple models
model = np.zeros((3, 2, 5))
jitter = np.zeros((3, 2, 5))
data = np.ones((3, 2))
errors = np.ones((3, 2))
seppa_indices = [np.array([1])]
chi2 = lnlike.chi2_lnlike(data, errors, None, model, jitter, seppa_indices)
assert chi2.shape == (3, 2, 5)
assert (chi2 == -0.5 * np.ones(
(3, 2, 5)) - np.log(np.sqrt(2 * np.pi * np.ones((3, 2, 5))))).all()
def test_chi2lnlike_withcov():
"""
Test the ability of ``orbitize.lnlike.chi2_lnlike()`` to work with some or all data having covariances
"""
### all all covariances
data = np.array([[5, -4], [3, -2], [1, 0]])
model = np.zeros(data.shape)
jitter = np.zeros(data.shape)
errs = np.array([[2, 2], [2, 2], [2, 2]])
covs = np.array([1, 0.25, 0.25])
corrs = covs / errs[:, 0] / errs[:, 1]
chi2s = lnlike.chi2_lnlike(data, errs, corrs, model, jitter, [])
residuals = data - model
for res, err, cov, chi2 in zip(residuals, errs, covs, chi2s):
cov_matrix = np.array([[err[0]**2, cov], [cov, err[1]**2]])
cov_inv = np.linalg.inv(cov_matrix)
cov_inv_dot_diff = np.dot(cov_inv, res)
logdet = np.linalg.slogdet(cov_matrix)[1]
res_cov_res = res.dot(cov_inv_dot_diff)
numpy_chi2 = -0.5 * (res_cov_res + logdet + 2 * np.log(2 * np.pi))
assert np.sum(chi2) == numpy_chi2
### only one covariance term
covs = np.array([1, np.nan, np.nan])
corrs = covs / errs[:, 0] / errs[:, 1]
new_chi2s = lnlike.chi2_lnlike(data, errs, corrs, model, jitter, [])
assert np.all(chi2s[0] == new_chi2s[0])