def test_arx(self):
arx = ARX(self.y, self.x, lags=3, hold_back=10, constant=False)
params = np.array([0.4, 0.3, 0.2, 1.0, 1.0])
data = arx.simulate(params, self.T, x=self.rng.randn(self.T + 500, 1))
assert isinstance(data, pd.DataFrame)
bounds = arx.bounds()
for b in bounds:
assert_equal(b[0], -np.inf)
assert_equal(b[1], np.inf)
assert_equal(len(bounds), 4)
assert_equal(arx.num_params, 4)
assert not arx.constant
a, b = arx.constraints()
assert_equal(a, np.empty((0, 4)))
assert_equal(b, np.empty(0))
res = arx.fit(last_obs=900, disp=DISPLAY)
assert res.fit_stop == 900
nobs = 900 - 10
rhs = np.zeros((nobs, 4))
y = self.y
lhs = y[10:900]
for i in range(10, 900):
rhs[i - 10, 0] = y[i - 1]
rhs[i - 10, 1] = y[i - 2]
rhs[i - 10, 2] = y[i - 3]
rhs[:, 3] = self.x[10:900, 0]
params = np.linalg.pinv(rhs).dot(lhs)
assert_almost_equal(params, res.params[:-1])
assert_equal(arx.hold_back, 10)
assert_equal(arx.lags, np.array([[1, 2, 3], [1, 2, 3]]))
assert_equal(arx.name, "AR-X")
assert_equal(arx.use_rotated, False)
assert isinstance(arx.__repr__(), str)
arx._repr_html_()
def test_ar(self):
ar = ARX(self.y, lags=3)
params = np.array([1.0, 0.4, 0.3, 0.2, 1.0])
data = ar.simulate(params, self.T)
assert len(data) == self.T
assert_equal(self.y, ar.y)
bounds = ar.bounds()
for b in bounds:
assert_equal(b[0], -np.inf)
assert_equal(b[1], np.inf)
assert_equal(len(bounds), 4)
assert_equal(ar.num_params, 4)
assert ar.constant
a, b = ar.constraints()
assert_equal(a, np.empty((0, 4)))
assert_equal(b, np.empty(0))
res = ar.fit(disp=DISPLAY)
nobs = 1000 - 3
rhs = np.ones((nobs, 4))
y = self.y
lhs = y[3:1000]
for i in range(3, 1000):
rhs[i - 3, 1] = y[i - 1]
rhs[i - 3, 2] = y[i - 2]
rhs[i - 3, 3] = y[i - 3]
params = np.linalg.pinv(rhs).dot(lhs)
assert_almost_equal(params, res.params[:-1])
forecasts = res.forecast(horizon=5)
direct = pd.DataFrame(index=np.arange(y.shape[0]),
columns=['h.' + str(i + 1) for i in range(5)],
dtype=np.float64)
params = res.params.iloc[:-1]
for i in range(2, y.shape[0]):
fcast = np.zeros(y.shape[0] + 5)
fcast[:y.shape[0]] = y.copy()
for h in range(1, 6):
reg = np.array([1.0, fcast[i + h - 1],
fcast[i + h - 2], fcast[i + h - 3]])
fcast[i + h] = reg.dot(params)
direct.iloc[i, :] = fcast[i + 1:i + 6]
assert isinstance(forecasts, ARCHModelForecast)
# TODO
# assert_frame_equal(direct, forecasts)
assert_equal(ar.hold_back, None)
assert_equal(ar.lags, np.array([[0, 1, 2], [1, 2, 3]]))
assert_equal(ar.name, 'AR')
assert_equal(ar.use_rotated, False)
ar.__repr__()
ar._repr_html_()
ar = ARX(self.y_df, lags=5)
ar.__repr__()
ar = ARX(self.y_series, lags=5)
ar.__repr__()
res = ar.fit(disp=DISPLAY)
assert isinstance(res.resid, pd.Series)
assert isinstance(res.conditional_volatility, pd.Series)
# Smoke bootstrap
summ = ar.fit(disp=DISPLAY).summary()
assert 'Constant Variance' in str(summ)
ar = ARX(self.y, lags=1, volatility=GARCH(), distribution=StudentsT())
res = ar.fit(disp=DISPLAY, update_freq=5, cov_type='mle')
res.param_cov
sims = res.forecast(horizon=5, method='simulation')
assert isinstance(sims.simulations.residual_variances, np.ndarray)
assert isinstance(sims.simulations.residuals, np.ndarray)
assert isinstance(sims.simulations.values, np.ndarray)
assert isinstance(sims.simulations.variances, np.ndarray)
def test_ar(self):
ar = ARX(self.y, lags=3)
params = np.array([1.0, 0.4, 0.3, 0.2, 1.0])
data = ar.simulate(params, self.T)
assert_equal(self.y, ar.y)
bounds = ar.bounds()
for b in bounds:
assert_equal(b[0], -np.inf)
assert_equal(b[1], np.inf)
assert_equal(len(bounds), 4)
assert_equal(ar.num_params, 4)
assert_true(ar.constant)
a, b = ar.constraints()
assert_equal(a, np.empty((0, 4)))
assert_equal(b, np.empty(0))
res = ar.fit()
nobs = 1000 - 3
rhs = np.ones((nobs, 4))
y = self.y
lhs = y[3:1000]
for i in range(3, 1000):
rhs[i - 3, 1] = y[i - 1]
rhs[i - 3, 2] = y[i - 2]
rhs[i - 3, 3] = y[i - 3]
params = np.linalg.pinv(rhs).dot(lhs)
assert_almost_equal(params, res.params[:-1])
forecasts = res.forecast(horizon=5)
direct = pd.DataFrame(index=np.arange(y.shape[0]),
columns=['h.' + str(i + 1) for i in range(5)],
dtype=np.float64)
params = res.params.iloc[:-1]
for i in range(2, y.shape[0]):
fcast = np.zeros(y.shape[0] + 5)
fcast[:y.shape[0]] = y.copy()
for h in range(1, 6):
reg = np.array(
[1.0, fcast[i + h - 1], fcast[i + h - 2], fcast[i + h - 3]])
fcast[i + h] = reg.dot(params)
direct.iloc[i, :] = fcast[i + 1:i + 6]
assert_frame_equal(direct, forecasts)
assert_equal(ar.first_obs, 3)
assert_equal(ar.last_obs, 1000)
assert_equal(ar.hold_back, None)
assert_equal(ar.lags, np.array([[0, 1, 2], [1, 2, 3]]))
assert_equal(ar.nobs, 997)
assert_equal(ar.name, 'AR')
assert_equal(ar.use_rotated, False)
ar.__repr__()
ar._repr_html_()
ar = ARX(self.y_df, lags=5)
ar.__repr__()
ar = ARX(self.y_series, lags=5)
ar.__repr__()
res = ar.fit()
assert_true(isinstance(res.resid, pd.Series))
assert_true(isinstance(res.conditional_volatility, pd.Series))
# Smoke tests
summ = ar.fit().summary()
ar = ARX(self.y, lags=1, volatility=GARCH(), distribution=StudentsT())
res = ar.fit(update_freq=5, cov_type='mle')
res.param_cov
res.plot()
res.plot(annualize='D')
res.plot(annualize='W')
res.plot(annualize='M')
assert_raises(ValueError, res.plot, annualize='unknown')
res.plot(scale=360)
res.hedgehog_plot()
def test_ar(self):
ar = ARX(self.y, lags=3)
params = np.array([1.0, 0.4, 0.3, 0.2, 1.0])
data = ar.simulate(params, self.T)
assert len(data) == self.T
assert_equal(self.y, ar.y)
bounds = ar.bounds()
for b in bounds:
assert_equal(b[0], -np.inf)
assert_equal(b[1], np.inf)
assert_equal(len(bounds), 4)
assert_equal(ar.num_params, 4)
assert ar.constant
a, b = ar.constraints()
assert_equal(a, np.empty((0, 4)))
assert_equal(b, np.empty(0))
res = ar.fit(disp=DISPLAY)
nobs = 1000 - 3
rhs = np.ones((nobs, 4))
y = self.y
lhs = y[3:1000]
for i in range(3, 1000):
rhs[i - 3, 1] = y[i - 1]
rhs[i - 3, 2] = y[i - 2]
rhs[i - 3, 3] = y[i - 3]
params = np.linalg.pinv(rhs).dot(lhs)
assert_almost_equal(params, res.params[:-1])
forecasts = res.forecast(horizon=5)
direct = pd.DataFrame(index=np.arange(y.shape[0]),
columns=['h.' + str(i + 1) for i in range(5)],
dtype=np.float64)
params = res.params.iloc[:-1]
for i in range(2, y.shape[0]):
fcast = np.zeros(y.shape[0] + 5)
fcast[:y.shape[0]] = y.copy()
for h in range(1, 6):
reg = np.array([1.0, fcast[i + h - 1],
fcast[i + h - 2], fcast[i + h - 3]])
fcast[i + h] = reg.dot(params)
direct.iloc[i, :] = fcast[i + 1:i + 6]
assert isinstance(forecasts, ARCHModelForecast)
# TODO
# assert_frame_equal(direct, forecasts)
assert_equal(ar.hold_back, None)
assert_equal(ar.lags, np.array([[0, 1, 2], [1, 2, 3]]))
assert_equal(ar.name, 'AR')
assert_equal(ar.use_rotated, False)
ar.__repr__()
ar._repr_html_()
ar = ARX(self.y_df, lags=5)
ar.__repr__()
ar = ARX(self.y_series, lags=5)
ar.__repr__()
res = ar.fit(disp=DISPLAY)
assert isinstance(res.resid, pd.Series)
assert isinstance(res.conditional_volatility, pd.Series)
# Smoke bootstrap
summ = ar.fit(disp=DISPLAY).summary()
assert 'Constant Variance' in str(summ)
ar = ARX(self.y, lags=1, volatility=GARCH(), distribution=StudentsT())
res = ar.fit(disp=DISPLAY, update_freq=5, cov_type='mle')
res.param_cov
def test_ar(self):
ar = ARX(self.y, lags=3)
params = np.array([1.0, 0.4, 0.3, 0.2, 1.0])
data = ar.simulate(params, self.T)
assert_equal(self.y, ar.y)
bounds = ar.bounds()
for b in bounds:
assert_equal(b[0], -np.inf)
assert_equal(b[1], np.inf)
assert_equal(len(bounds), 4)
assert_equal(ar.num_params, 4)
assert_true(ar.constant)
a, b = ar.constraints()
assert_equal(a, np.empty((0, 4)))
assert_equal(b, np.empty(0))
res = ar.fit()
nobs = 1000 - 3
rhs = np.ones((nobs, 4))
y = self.y
lhs = y[3:1000]
for i in range(3, 1000):
rhs[i - 3, 1] = y[i - 1]
rhs[i - 3, 2] = y[i - 2]
rhs[i - 3, 3] = y[i - 3]
params = np.linalg.pinv(rhs).dot(lhs)
assert_almost_equal(params, res.params[:-1])
forecasts = res.forecast(horizon=5)
direct = pd.DataFrame(index=np.arange(y.shape[0]),
columns=['h.' + str(i + 1) for i in range(5)],
dtype=np.float64)
params = res.params.iloc[:-1]
for i in range(2, y.shape[0]):
fcast = np.zeros(y.shape[0] + 5)
fcast[:y.shape[0]] = y.copy()
for h in range(1, 6):
reg = np.array([
1.0, fcast[i + h - 1], fcast[i + h - 2], fcast[i + h - 3]
])
fcast[i + h] = reg.dot(params)
direct.iloc[i, :] = fcast[i + 1:i + 6]
assert_frame_equal(direct, forecasts)
assert_equal(ar.first_obs, 3)
assert_equal(ar.last_obs, 1000)
assert_equal(ar.hold_back, None)
assert_equal(ar.lags, np.array([[0, 1, 2], [1, 2, 3]]))
assert_equal(ar.nobs, 997)
assert_equal(ar.name, 'AR')
assert_equal(ar.use_rotated, False)
ar.__repr__()
ar._repr_html_()
ar = ARX(self.y_df, lags=5)
ar.__repr__()
ar = ARX(self.y_series, lags=5)
ar.__repr__()
res = ar.fit()
assert_true(isinstance(res.resid, pd.Series))
assert_true(isinstance(res.conditional_volatility, pd.Series))
# Smoke tests
summ = ar.fit().summary()
ar = ARX(self.y, lags=1, volatility=GARCH(), distribution=StudentsT())
res = ar.fit(update_freq=5, cov_type='mle')
res.param_cov
res.plot()
res.plot(annualize='D')
res.plot(annualize='W')
res.plot(annualize='M')
assert_raises(ValueError, res.plot, annualize='unknown')
res.plot(scale=360)
res.hedgehog_plot()
