def test_log_likelihood(self):
model = GARCHModel(0.2, 0.3, 0.4, sc=self.sc)
n = 10000
ts = np.array(model.sample(n))
logLikelihoodWithRightModel = model.log_likelihood(ts)
logLikelihoodWithWrongModel1 = GARCHModel(
0.3, 0.4, 0.5, sc=self.sc).log_likelihood(ts)
logLikelihoodWithWrongModel2 = GARCHModel(
0.25, 0.35, 0.45, sc=self.sc).log_likelihood(ts)
logLikelihoodWithWrongModel3 = GARCHModel(
0.1, 0.2, 0.3, sc=self.sc).log_likelihood(ts)
self.assertTrue(
logLikelihoodWithRightModel > logLikelihoodWithWrongModel1)
self.assertTrue(
logLikelihoodWithRightModel > logLikelihoodWithWrongModel2)
self.assertTrue(
logLikelihoodWithRightModel > logLikelihoodWithWrongModel3)
self.assertTrue(
logLikelihoodWithWrongModel2 > logLikelihoodWithWrongModel1)
def test_gradient(self):
alpha = 0.3
beta = 0.4
omega = 0.2
genModel = GARCHModel(omega, alpha, beta, sc=self.sc)
n = 10000
ts = np.array(genModel.sample(n))
gradient1 = GARCHModel(omega + 0.1,
alpha + 0.05,
beta + 0.1,
sc=self.sc).gradient(ts)
for g in gradient1:
self.assertTrue(g < 0.0)
gradient2 = GARCHModel(omega - 0.1,
alpha - 0.05,
beta - 0.1,
sc=self.sc).gradient(ts)
for g in gradient2:
self.assertTrue(g > 0.0)
