def test_stationary_solve_1d():
b = np.random.uniform(size=10)
r = np.random.uniform(size=9)
t = np.concatenate((np.r_[1], r))
tmat = toeplitz(t)
soln = np.linalg.solve(tmat, b)
soln1 = linalg.stationary_solve(r, b)
assert_allclose(soln, soln1, rtol=1e-5, atol=1e-5)
def test_stationary_solve_1d():
b = np.random.uniform(size=10)
r = np.random.uniform(size=9)
t = np.concatenate((np.r_[1], r))
tmat = toeplitz(t)
soln = np.linalg.solve(tmat, b)
soln1 = linalg.stationary_solve(r, b)
assert_allclose(soln, soln1, rtol=1e-5, atol=1e-5)
def covariance_matrix_solve(self, expval, index, stdev, rhs):
if self.grid == False:
return super(Stationary, self).covariance_matrix_solve(expval, index, stdev, rhs)
from statsmodels.tools.linalg import stationary_solve
r = np.zeros(len(expval))
r[0:self.max_lag] = self.dep_params
return [stationary_solve(r, x) for x in rhs]
def covariance_matrix_solve(self, expval, index, stdev, rhs):
if not self.grid:
return super(Stationary,
self).covariance_matrix_solve(expval, index, stdev,
rhs)
from statsmodels.tools.linalg import stationary_solve
r = np.zeros(len(expval))
r[0:self.max_lag] = self.dep_params[1:]
rslt = []
for x in rhs:
if x.ndim == 1:
y = x / stdev
rslt.append(stationary_solve(r, y) / stdev)
else:
y = x / stdev[:, None]
rslt.append(stationary_solve(r, y) / stdev[:, None])
return rslt
