def test_CopulaGaussian(self):
df_corr = pd.DataFrame(dict(var1=["x"], var2=["y"], corr=[0.5]))
Sigma_h = np.linalg.cholesky(np.array([[1.0, 0.5], [0.5, 1.0]]))
copula = gr.CopulaGaussian(["x", "y"], df_corr=df_corr)
df_res = copula.sample(seed=101)
self.assertTrue(np.isclose(copula.Sigma_h, Sigma_h).all)
self.assertTrue(set(df_res.columns) == set(["x", "y"]))
## Test raises
df_corr_invalid = pd.DataFrame(
dict(var1=["x", "x"], var2=["y", "z"], corr=[0, 0])
)
with self.assertRaises(ValueError):
gr.CopulaGaussian(["x", "y"], df_corr=df_corr_invalid)
## Transforms invariant
z = np.array([0, 0])
u = copula.z2u(z)
zp = copula.u2z(u)
self.assertTrue(np.all(z == zp))
## Jacobian accurate
dudz_fd = np.zeros((2, 2))
dudz_fd[0, :] = (copula.z2u(z + np.array([h, 0])) - copula.z2u(z)) / h
dudz_fd[1, :] = (copula.z2u(z + np.array([0, h])) - copula.z2u(z)) / h
dudz_p = copula.dudz(z)
self.assertTrue(np.allclose(dudz_fd, dudz_p))
def setUp(self):
self.density = gr.Density(
marginals=dict(
x=gr.MarginalNamed(d_name="uniform",
d_param={
"loc": -1,
"scale": 2
}),
y=gr.MarginalNamed(d_name="uniform",
d_param={
"loc": -1,
"scale": 2
}),
),
copula=gr.CopulaGaussian(["x", "y"],
pd.DataFrame(
dict(var1=["x"],
var2=["y"],
corr=[0.5]))),
)
self.density_ind = gr.Density(
marginals=dict(
x=gr.MarginalNamed(d_name="uniform",
d_param={
"loc": -1,
"scale": 2
}),
y=gr.MarginalNamed(d_name="norm",
d_param={
"loc": 0,
"scale": 1
}),
),
copula=gr.CopulaIndependence(["x", "y"]),
)
self.density_gauss = gr.Density(
marginals=dict(
x=gr.MarginalNamed(d_name="uniform",
d_param={
"loc": -1,
"scale": 2
}),
y=gr.MarginalNamed(d_name="norm",
d_param={
"loc": 0,
"scale": 1
}),
),
copula=gr.CopulaGaussian(["x", "y"],
pd.DataFrame(
dict(var1=["x"],
var2=["y"],
corr=[0.5]))),
)
def setUp(self):
self.density = gr.Density(
marginals=dict(
x=gr.MarginalNamed(d_name="uniform", d_param={"loc": -1, "scale": 2}),
y=gr.MarginalNamed(d_name="uniform", d_param={"loc": -1, "scale": 2}),
),
copula=gr.CopulaGaussian(
pd.DataFrame(dict(var1=["x"], var2=["y"], corr=[0.5]))
),
)