def test_reorder_vector():
nobs = 5
k_endog = 3
missing = np.zeros((k_endog, nobs))
missing[0, 0] = 1
missing[:2, 1] = 1
missing[0, 2] = 1
missing[2, 2] = 1
missing[1, 3] = 1
missing[2, 4] = 1
given = np.zeros((k_endog, nobs))
given[:, :] = np.array([1, 2, 3])[:, np.newaxis]
desired = given.copy()
given[:, 0] = [2, 3, 0]
desired[:, 0] = [0, 2, 3]
given[:, 1] = [3, 0, 0]
desired[:, 1] = [0, 0, 3]
given[:, 2] = [2, 0, 0]
desired[:, 2] = [0, 2, 0]
given[:, 3] = [1, 3, 0]
desired[:, 3] = [1, 0, 3]
given[:, 4] = [1, 2, 0]
desired[:, 4] = [1, 2, 0]
actual = np.asfortranarray(given.copy())
missing = np.asfortranarray(missing.astype(np.int32))
tools.reorder_missing_vector(actual, missing, inplace=True)
assert_equal(actual, desired)
def test_reorder_vector():
nobs = 5
k_endog = 3
missing = np.zeros((k_endog, nobs))
missing[0, 0] = 1
missing[:2, 1] = 1
missing[0, 2] = 1
missing[2, 2] = 1
missing[1, 3] = 1
missing[2, 4] = 1
given = np.zeros((k_endog, nobs))
given[:, :] = np.array([1, 2, 3])[:, np.newaxis]
desired = given.copy()
given[:, 0] = [2, 3, 0]
desired[:, 0] = [0, 2, 3]
given[:, 1] = [3, 0, 0]
desired[:, 1] = [0, 0, 3]
given[:, 2] = [2, 0, 0]
desired[:, 2] = [0, 2, 0]
given[:, 3] = [1, 3, 0]
desired[:, 3] = [1, 0, 3]
given[:, 4] = [1, 2, 0]
desired[:, 4] = [1, 2, 0]
actual = np.asfortranarray(given.copy())
missing = np.asfortranarray(missing.astype(np.int32))
tools.reorder_missing_vector(actual, missing, inplace=True)
assert_equal(actual, desired)
def update_smoother(self, smoother):
"""
Update the smoother results
Parameters
----------
smoother : KalmanSmoother
The model object from which to take the updated values.
Notes
-----
This method is rarely required except for internal usage.
"""
# Copy the appropriate output
attributes = []
# Since update_representation will already have been called, we can
# use the boolean options smoother_* and know they match the smoother
# itself
if self.smoother_state or self.smoother_disturbance:
attributes.append('scaled_smoothed_estimator')
if self.smoother_state_cov or self.smoother_disturbance_cov:
attributes.append('scaled_smoothed_estimator_cov')
if self.smoother_state:
attributes.append('smoothed_state')
if self.smoother_state_cov:
attributes.append('smoothed_state_cov')
if self.smoother_state_autocov:
attributes.append('smoothed_state_autocov')
if self.smoother_disturbance:
attributes += [
'smoothing_error',
'smoothed_measurement_disturbance',
'smoothed_state_disturbance'
]
if self.smoother_disturbance_cov:
attributes += [
'smoothed_measurement_disturbance_cov',
'smoothed_state_disturbance_cov'
]
has_missing = np.sum(self.nmissing) > 0
for name in self._smoother_attributes:
if name == 'smoother_output':
pass
elif name in attributes:
if name in ['smoothing_error',
'smoothed_measurement_disturbance']:
vector = getattr(smoother, name, None)
if vector is not None and has_missing:
vector = np.array(reorder_missing_vector(
vector, self.missing, prefix=self.prefix))
else:
vector = np.array(vector, copy=True)
setattr(self, name, vector)
elif name == 'smoothed_measurement_disturbance_cov':
matrix = getattr(smoother, name, None)
if matrix is not None and has_missing:
matrix = reorder_missing_matrix(
matrix, self.missing, reorder_rows=True,
reorder_cols=True, prefix=self.prefix)
# In the missing data case, we want to set the missing
# components equal to their unconditional distribution
copy_index_matrix(
self.obs_cov, matrix, self.missing,
index_rows=True, index_cols=True, inplace=True,
prefix=self.prefix)
else:
matrix = np.array(matrix, copy=True)
setattr(self, name, matrix)
else:
setattr(self, name,
np.array(getattr(smoother, name, None), copy=True))
else:
setattr(self, name, None)
# Diffuse objects
self.scaled_smoothed_diffuse_estimator = None
self.scaled_smoothed_diffuse1_estimator_cov = None
self.scaled_smoothed_diffuse2_estimator_cov = None
if self.nobs_diffuse > 0:
self.scaled_smoothed_diffuse_estimator = np.array(
smoother.scaled_smoothed_diffuse_estimator, copy=True)
self.scaled_smoothed_diffuse1_estimator_cov = np.array(
smoother.scaled_smoothed_diffuse1_estimator_cov, copy=True)
self.scaled_smoothed_diffuse2_estimator_cov = np.array(
smoother.scaled_smoothed_diffuse2_estimator_cov, copy=True)
# Adjustments
# For r_t (and similarly for N_t), what was calculated was
# r_T, ..., r_{-1}. We only want r_0, ..., r_T
# so exclude the appropriate element so that the time index is
# consistent with the other returned output
# r_t stored such that scaled_smoothed_estimator[0] == r_{-1}
start = 1
end = None
if 'scaled_smoothed_estimator' in attributes:
self.scaled_smoothed_estimator = (
self.scaled_smoothed_estimator[:, start:end]
)
if 'scaled_smoothed_estimator_cov' in attributes:
self.scaled_smoothed_estimator_cov = (
self.scaled_smoothed_estimator_cov[:, :, start:end]
)
# Clear the smoothed forecasts
self._smoothed_forecasts = None
self._smoothed_forecasts_error = None
self._smoothed_forecasts_error_cov = None
def update_smoother(self, smoother):
"""
Update the smoother results
Parameters
----------
smoother : KalmanSmoother
The model object from which to take the updated values.
Notes
-----
This method is rarely required except for internal usage.
"""
# Copy the appropriate output
attributes = []
# Since update_representation will already have been called, we can
# use the boolean options smoother_* and know they match the smoother
# itself
if self.smoother_state or self.smoother_disturbance:
attributes.append('scaled_smoothed_estimator')
if self.smoother_state_cov or self.smoother_disturbance_cov:
attributes.append('scaled_smoothed_estimator_cov')
if self.smoother_state:
attributes.append('smoothed_state')
if self.smoother_state_cov:
attributes.append('smoothed_state_cov')
if self.smoother_state_autocov:
attributes.append('smoothed_state_autocov')
if self.smoother_disturbance:
attributes += [
'smoothing_error',
'smoothed_measurement_disturbance',
'smoothed_state_disturbance'
]
if self.smoother_disturbance_cov:
attributes += [
'smoothed_measurement_disturbance_cov',
'smoothed_state_disturbance_cov'
]
has_missing = np.sum(self.nmissing) > 0
for name in self._smoother_attributes:
if name == 'smoother_output':
pass
elif name in attributes:
if name in ['smoothing_error',
'smoothed_measurement_disturbance']:
vector = getattr(smoother, name, None)
if (not self._compatibility_mode and vector is not None and
has_missing):
vector = np.array(reorder_missing_vector(
vector, self.missing, prefix=self.prefix))
else:
vector = np.array(vector, copy=True)
setattr(self, name, vector)
elif name == 'smoothed_measurement_disturbance_cov':
matrix = getattr(smoother, name, None)
if (not self._compatibility_mode and matrix is not None and
has_missing):
matrix = reorder_missing_matrix(
matrix, self.missing, reorder_rows=True,
reorder_cols=True, prefix=self.prefix)
# In the missing data case, we want to set the missing
# components equal to their unconditional distribution
copy_index_matrix(
self.obs_cov, matrix, self.missing,
index_rows=True, index_cols=True, inplace=True,
prefix=self.prefix)
else:
matrix = np.array(matrix, copy=True)
setattr(self, name, matrix)
else:
setattr(self, name,
np.array(getattr(smoother, name, None), copy=True))
else:
setattr(self, name, None)
# Adjustments
# For r_t (and similarly for N_t), what was calculated was
# r_T, ..., r_{-1}. We only want r_0, ..., r_T
# so exclude the appropriate element so that the time index is
# consistent with the other returned output
if not self._compatibility_mode:
# r_t stored such that scaled_smoothed_estimator[0] == r_{-1}
start = 1
end = None
else:
# r_t was stored such that scaled_smoothed_estimator[-1] == r_{-1}
start = None
end = -1
if 'scaled_smoothed_estimator' in attributes:
self.scaled_smoothed_estimator = (
self.scaled_smoothed_estimator[:, start:end]
)
if 'scaled_smoothed_estimator_cov' in attributes:
self.scaled_smoothed_estimator_cov = (
self.scaled_smoothed_estimator_cov[:, :, start:end]
)
# Clear the smoothed forecasts
self._smoothed_forecasts = None
self._smoothed_forecasts_error = None
self._smoothed_forecasts_error_cov = None