def test_seeded_run_ar(self):
random.seed(1)
m = ARIMA(c=1, sigma=1, pcoeff=[1, -1, 0.5])
x = m.generate(10)
expected = [
2.28818475, 4.73763036, 3.51578142, 0.15769978, -1.08143967,
1.55008577, 2.68827216, 0.16063711, -0.55328019, 1.76359339
]
assert x == pytest.approx(expected, rel=1e-5)
def test_seeded_run_ma(self):
random.seed(1)
# see note on qcoeff signum on TestMA.test_seeded_run
m = ARIMA(c=0, sigma=1, qcoeff=[-1, -2, -3])
x = m.generate(10)
expected = [
1.28818475, 2.73763036, 4.09215092, 6.06523763, 2.62429158,
-2.39091857, -5.46874604, -5.67278323, -3.18772026, -5.60728182
]
assert x == pytest.approx(expected, rel=1e-5)
def test_seeded_random_walk_as_arima(self):
random.seed(1)
m = ARIMA(c=0, sigma=1, d=1)
x = m.generate(10)
expected = [
1.2881847531554629, 2.737630361855234, 2.8039661707934957,
2.039422519821864, 0.9472493047177224, 0.9785838215494392,
-0.04351934846143368, -1.4803487935639636, -1.2810368170802098,
-1.147662212421605
]
assert x == pytest.approx(expected, rel=1e-5)
def test_full_arima(self):
random.seed(1)
# see note on qcoeff signum on TestMA.test_seeded_run
m = ARIMA(pcoeff=[1, -1], d=2, qcoeff=[-1, -1, -1, -1, -1])
x = m.generate(10)
expected = [
1.2881847531554629, 6.602184621321622, 17.457781022136512,
31.868581360411298, 45.24023840821498, 54.99613204963708,
60.84370147353184, 62.18073093086161, 58.83054165128004,
52.11658886910866
]
assert x == pytest.approx(expected, rel=1e-5)
def test_seeded_run_ma_using_buffer(self):
random.seed(1)
# see note on qcoeff signum on TestMA.test_seeded_run
m = ARIMA(c=0, sigma=1, qcoeff=[-1, -2, -3], e_buff=[10, 9, 8])
x = m.generate(10)
expected = [
57.28818475315546, 45.73763036185523, 28.09215092394896,
6.0652376348325605, 2.6242915779000637, -2.390918573400903,
-5.468746036302335, -5.672783226762393, -3.1877202581453714,
-5.60728181909532
]
assert x == pytest.approx(expected, rel=1e-5)
def test_seeded_run_arma_using_buffers(self):
random.seed(1)
# see note on qcoeff signum on TestMA.test_seeded_run
m = ARIMA(c=1,
sigma=1,
pcoeff=[1, -1, 0.5],
qcoeff=[-1, -2, 1],
x_buff=[-4, -6, 8],
e_buff=[2, 1, -1])
x = m.generate(10)
expected = [
13.288184753155463, 3.0258151150106967, -0.1702187141958067,
5.360557169581937, 4.870192584384071, -2.2317357514497194,
-5.832221183458628, -1.469479506162701, 0.9338154760988293,
-1.031684748519612
]
assert x == pytest.approx(expected, rel=1e-5)
def test_full_arima_10_equivalent_to_5_and_5(self):
random.seed(1)
# see note on qcoeff signum on TestMA.test_seeded_run
m1 = ARIMA(pcoeff=[1, -1], d=2, qcoeff=[-1, -1, -1, -1, -1])
x1 = m1.generate(10)
random.seed(1)
m2 = ARIMA(pcoeff=[1, -1], d=2, qcoeff=[-1, -1, -1, -1, -1])
x2 = m2.generate(5) + m2.generate(5)
assert x1 == pytest.approx(x2, rel=1e-5)
def test_white_noise(self):
m = ARIMA(c=0, sigma=1, pcoeff=[])
s = pd.Series(m.generate(1000))
assert s.std() == pytest.approx(1.0, abs=1e2)
assert s.mean() == pytest.approx(0.0, abs=1e2)
def test_periodic(self):
m = ARIMA(c=1, sigma=0, pcoeff=[1, -1])
x = m.generate(10)
assert x == [1.0, 2.0, 2.0, 1.0, 0.0, 0.0, 1.0, 2.0, 2.0, 1.0]
def test_linear_using_buffer(self):
m = ARIMA(c=1, sigma=0, pcoeff=[1], x_buff=[4])
x = m.generate(10)
assert x == [float(x) for x in range(5, 15)]
def test_linear_multiple_calls(self):
m = ARIMA(c=1, sigma=0, pcoeff=[1])
x = m.generate(10)
assert x == [float(x) for x in range(1, 11)]
x = m.generate(10)
assert x == [float(x) for x in range(11, 21)]
def test_linear(self):
m = ARIMA(c=1, sigma=0, pcoeff=[1])
x = m.generate(10)
assert x == [float(x) for x in range(1, 11)]
def test_constant_generation(self):
m = ARIMA(c=1, sigma=0)
x = m.generate(10)
assert x == [1] * 10
def test_default_creation(self):
m = ARIMA()
assert m