def test_combine_subset_regression(self):
# split sample into two, use first sample as prior for second
endog = self.endog
exog = self.exog
nobs = len(endog)
n05 = nobs // 2
np.random.seed(987125)
# shuffle to get random subsamples
shuffle_idx = np.random.permutation(np.arange(nobs))
ys = endog[shuffle_idx]
xs = exog[shuffle_idx]
k = 10
res_ols0 = OLS(ys[:n05], xs[:n05, :k]).fit()
res_ols1 = OLS(ys[n05:], xs[n05:, :k]).fit()
w = res_ols1.scale / res_ols0.scale #1.01
mod_1 = TheilGLS(ys[n05:], xs[n05:, :k], r_matrix=np.eye(k), q_matrix=res_ols0.params, sigma_prior=w * res_ols0.cov_params())
res_1p = mod_1.fit(cov_type='data-prior')
res_1s = mod_1.fit(cov_type='sandwich')
res_olsf = OLS(ys, xs[:, :k]).fit()
assert_allclose(res_1p.params, res_olsf.params, rtol=1e-9)
corr_fact = 0.96156318 # corrct for differences in scale computation
assert_allclose(res_1p.bse, res_olsf.bse * corr_fact, rtol=1e-3)
# regression test, does not verify numbers
# especially why are these smaller than OLS on full sample
# in larger sample, nobs=600, those were close to full OLS
bse1 = np.array([
0.27609914, 0.15808869, 0.39880789, 0.78583194, 0.68619331,
0.56252314, 0.55757562, 0.68538523, 0.39695081, 0.55988991])
assert_allclose(res_1s.bse, bse1, rtol=1e-7)
def _initialize(cls):
y, x = cls.y, cls.x
# adding 10 to avoid strict rtol at predicted values close to zero
y = y + 10
k = x.shape[1]
cov_type = 'HC0'
restriction = np.eye(k)[2:]
modp = TheilGLS(y, x, r_matrix=restriction)
# the corresponding Theil penweight seems to be 2 * nobs / sigma2_e
cls.res2 = modp.fit(pen_weight=120.74564413221599 * 1000)
pen = smpen.L2ContraintsPenalty(restriction=restriction)
mod = GLMPenalized(y, x, family=family.Gaussian(),
penal=pen)
# use default weight for GLMPenalized
mod.pen_weight *= 1
cls.res1 = mod.fit(cov_type=cov_type, method='bfgs', maxiter=100,
disp=0, trim=False)
cls.k_nonzero = k
cls.exog_index = slice(None, cls.k_nonzero, None)
cls.k_params = x.shape[1]
cls.atol = 1e-5
cls.rtol = 1e-5
def setup_class(cls):
y, x = cls.get_sample()
nobs = len(y)
weights = (np.arange(nobs) < (nobs // 2)) + 0.5
mod1 = TheilGLS(y, x, sigma=weights, sigma_prior=[0, 0, 1., 1.])
cls.res1 = mod1.fit(200000)
cls.res2 = GLS(y, x[:, :3], sigma=weights).fit()
def setup_class(cls):
y, x = cls.get_sample()
nobs = len(y)
weights = (np.arange(nobs) < (nobs // 2)) + 0.5
mod1 = TheilGLS(y, x, sigma=weights, sigma_prior=[0, 0, 1., 1.])
cls.res1 = mod1.fit(200000)
cls.res2 = GLS(y, x[:, :3], sigma=weights).fit()
def _initialize(cls):
y, x = cls.y, cls.x
# adding 10 to avoid strict rtol at predicted values close to zero
y = y + 10
k = x.shape[1]
cov_type = 'HC0'
restriction = np.eye(k)[2:]
modp = TheilGLS(y, x, r_matrix=restriction)
# the corresponding Theil penweight seems to be 2 * nobs / sigma2_e
cls.res2 = modp.fit(pen_weight=120.74564413221599 * 1000, use_t=False)
pen = smpen.L2ContraintsPenalty(restriction=restriction)
mod = GLMPenalized(y, x, family=family.Gaussian(), penal=pen)
# use default weight for GLMPenalized
mod.pen_weight *= 1
cls.res1 = mod.fit(cov_type=cov_type,
method='bfgs',
maxiter=100,
disp=0,
trim=False)
cls.k_nonzero = k
cls.exog_index = slice(None, cls.k_nonzero, None)
cls.k_params = x.shape[1]
cls.atol = 1e-5
cls.rtol = 1e-5
def setup_class(cls):
cls.skip_inference = True
y, x = cls.get_sample()
xd = np.column_stack((x, x))
#sp = np.zeros(5), np.ones(5)
r_matrix = np.eye(5, 10, 5)
mod1 = TheilGLS(y, xd, r_matrix=r_matrix) #sigma_prior=[0, 0, 1., 1.])
cls.res1 = mod1.fit(0.001, cov_type='data-prior')
cls.res2 = OLS(y, x).fit()
def setup_class(cls):
cls.skip_inference = True
y, x = cls.get_sample()
xd = np.column_stack((x, x))
#sp = np.zeros(5), np.ones(5)
r_matrix = np.eye(5, 10, 5)
mod1 = TheilGLS(y, xd, r_matrix=r_matrix) #sigma_prior=[0, 0, 1., 1.])
cls.res1 = mod1.fit(0.001, cov_type='data-prior')
cls.res2 = OLS(y, x).fit()
def setup_class(cls):
y, x = cls.get_sample()
#merge var1 and var2
x2 = x[:, :2].copy()
x2[:, 1] += x[:, 2]
#mod1 = TheilGLS(y, x, r_matrix =[[0, 1, -1, 0, 0]])
mod1 = TheilGLS(y, x[:, :3], r_matrix=[[0, 1, -1]])
cls.res1 = mod1.fit(200000)
cls.res2 = OLS(y, x2).fit()
# adjust precision, careful: cls.tol is mutable
tol = {'pvalues': (1e-4, 2e-7), 'tvalues': (5e-4, 0)}
tol.update(cls.tol)
cls.tol = tol
def setup_class(cls):
exog, penalty_matrix, restriction = cls._init()
endog = data_mcycle['accel']
modp = TheilGLS(endog, exog, r_matrix=restriction)
# scaling of penweith in R mgcv
s_scale_r = 0.02630734
# Theil penweight uses preliminary sigma2_e to scale penweight
sigma_e = 1405.7950179165323
cls.pw = pw = 1 / sigma_e / s_scale_r
cls.res1 = modp.fit(pen_weight=pw, cov_type=cls.cov_type)
cls.res2 = results_pls.pls5
cls.rtol_fitted = 1e-7
cls.covp_corrfact = 0.99786932844203202
def setup_class(cls):
exog, penalty_matrix, restriction = cls._init()
endog = data_mcycle['accel']
modp = TheilGLS(endog, exog, r_matrix=restriction)
# scaling of penweith in R mgcv
s_scale_r = 0.02630734
# Theil penweight uses preliminary sigma2_e to scale penweight
sigma_e = 1405.7950179165323
cls.pw = pw = 1 / sigma_e / s_scale_r
cls.res1 = modp.fit(pen_weight=pw, cov_type=cls.cov_type)
cls.res2 = results_pls.pls5
cls.rtol_fitted = 1e-7
cls.covp_corrfact = 0.99786932844203202
def setup_class(cls):
y, x = cls.get_sample()
#merge var1 and var2
x2 = x[:, :2].copy()
x2[:, 1] += x[:, 2]
#mod1 = TheilGLS(y, x, r_matrix =[[0, 1, -1, 0, 0]])
mod1 = TheilGLS(y, x[:, :3], r_matrix =[[0, 1, -1]])
cls.res1 = mod1.fit(200000)
cls.res2 = OLS(y, x2).fit()
# adjust precision, careful: cls.tol is mutable
tol = {'pvalues': (1e-4, 2e-7),
'tvalues': (5e-4, 0)}
tol.update(cls.tol)
cls.tol = tol
def setup_class(cls):
cur_dir = os.path.dirname(os.path.abspath(__file__))
filepath = os.path.join(cur_dir, "results",
"theil_textile_predict.csv")
cls.res_predict = pd.read_csv(filepath, sep=",")
names = "year lconsump lincome lprice".split()
data = np.array(
'''\
1923 1.99651 1.98543 2.00432
1924 1.99564 1.99167 2.00043
1925 2 2 2
1926 2.04766 2.02078 1.95713
1927 2.08707 2.02078 1.93702
1928 2.07041 2.03941 1.95279
1929 2.08314 2.04454 1.95713
1930 2.13354 2.05038 1.91803
1931 2.18808 2.03862 1.84572
1932 2.18639 2.02243 1.81558
1933 2.20003 2.00732 1.78746
1934 2.14799 1.97955 1.79588
1935 2.13418 1.98408 1.80346
1936 2.22531 1.98945 1.72099
1937 2.18837 2.0103 1.77597
1938 2.17319 2.00689 1.77452
1939 2.2188 2.0162 1.78746'''.split(), float).reshape(-1, 4)
endog = data[:, 1]
# constant at the end to match Stata
exog = np.column_stack((data[:, 2:], np.ones(endog.shape[0])))
#prior(lprice -0.7 0.15 lincome 1 0.15) cov(lprice lincome -0.01)
r_matrix = np.array([[1, 0, 0], [0, 1, 0]])
r_mean = [1, -0.7]
cov_r = np.array([[0.15**2, -0.01], [-0.01, 0.15**2]])
mod = TheilGLS(endog,
exog,
r_matrix,
q_matrix=r_mean,
sigma_prior=cov_r)
cls.res1 = mod.fit(cov_type='data-prior', use_t=True)
#cls.res1._cache['scale'] = 0.0001852252884817586 # from tg_mixed
cls.res1._cache['scale'] = 0.00018334123641580062 # from OLS
from .results import results_theil_textile as resmodule
cls.res2 = resmodule.results_theil_textile
def setup_class(cls):
cur_dir = os.path.dirname(os.path.abspath(__file__))
filepath = os.path.join(cur_dir, "results",
"theil_textile_predict.csv")
cls.res_predict = pd.read_csv(filepath, sep=",")
names = "year lconsump lincome lprice".split()
data = np.array('''\
1923 1.99651 1.98543 2.00432
1924 1.99564 1.99167 2.00043
1925 2 2 2
1926 2.04766 2.02078 1.95713
1927 2.08707 2.02078 1.93702
1928 2.07041 2.03941 1.95279
1929 2.08314 2.04454 1.95713
1930 2.13354 2.05038 1.91803
1931 2.18808 2.03862 1.84572
1932 2.18639 2.02243 1.81558
1933 2.20003 2.00732 1.78746
1934 2.14799 1.97955 1.79588
1935 2.13418 1.98408 1.80346
1936 2.22531 1.98945 1.72099
1937 2.18837 2.0103 1.77597
1938 2.17319 2.00689 1.77452
1939 2.2188 2.0162 1.78746'''.split(), float).reshape(-1, 4)
endog = data[:, 1]
# constant at the end to match Stata
exog = np.column_stack((data[:, 2:], np.ones(endog.shape[0])))
#prior(lprice -0.7 0.15 lincome 1 0.15) cov(lprice lincome -0.01)
r_matrix = np.array([[1, 0, 0], [0, 1, 0]])
r_mean = [1, -0.7]
cov_r = np.array([[0.15**2, -0.01], [-0.01, 0.15**2]])
mod = TheilGLS(endog, exog, r_matrix, q_matrix=r_mean, sigma_prior=cov_r)
cls.res1 = mod.fit(cov_type='data-prior')
#cls.res1._cache['scale'] = 0.0001852252884817586 # from tg_mixed
cls.res1._cache['scale'] = 0.00018334123641580062 # from OLS
from .results import results_theil_textile as resmodule
cls.res2 = resmodule.results_theil_textile
def setup_class(cls):
y, x = cls.get_sample()
#merge var1 and var2
x2 = x[:, :2].copy()
x2[:, 1] += x[:, 2]
#mod1 = TheilGLS(y, x, r_matrix =[[0, 1, -1, 0, 0]])
mod1 = TheilGLS(y, x[:, :3], r_matrix =[[0, 1, -1]])
cls.res1 = mod1.fit(100)
cls.res2 = OLS(y, x2).fit()
# adjust precision, careful: cls.tol is mutable
import copy
tol = copy.copy(cls.tol)
tol2 = {'default': (0.15, 0),
'params': (0.05, 0),
'pvalues': (0.02, 0.001),
}
tol.update(tol2)
cls.tol = tol
def setup_class(cls):
y, x = cls.get_sample()
#merge var1 and var2
x2 = x[:, :2].copy()
x2[:, 1] += x[:, 2]
#mod1 = TheilGLS(y, x, r_matrix =[[0, 1, -1, 0, 0]])
mod1 = TheilGLS(y, x[:, :3], r_matrix =[[0, 1, -1]])
cls.res1 = mod1.fit(100)
cls.res2 = OLS(y, x2).fit()
# adjust precision, careful: cls.tol is mutable
import copy
tol = copy.copy(cls.tol)
tol2 = {'default': (0.15, 0),
'params': (0.05, 0),
'pvalues': (0.02, 0.001),
}
tol.update(tol2)
cls.tol = tol
def test_combine_subset_regression(self):
# split sample into two, use first sample as prior for second
endog = self.endog
exog = self.exog
nobs = len(endog)
n05 = nobs // 2
np.random.seed(987125)
# shuffle to get random subsamples
shuffle_idx = np.random.permutation(np.arange(nobs))
ys = endog[shuffle_idx]
xs = exog[shuffle_idx]
k = 10
res_ols0 = OLS(ys[:n05], xs[:n05, :k]).fit()
res_ols1 = OLS(ys[n05:], xs[n05:, :k]).fit()
w = res_ols1.scale / res_ols0.scale #1.01
mod_1 = TheilGLS(ys[n05:],
xs[n05:, :k],
r_matrix=np.eye(k),
q_matrix=res_ols0.params,
sigma_prior=w * res_ols0.cov_params())
res_1p = mod_1.fit(cov_type='data-prior')
res_1s = mod_1.fit(cov_type='sandwich')
res_olsf = OLS(ys, xs[:, :k]).fit()
assert_allclose(res_1p.params, res_olsf.params, rtol=1e-9)
corr_fact = np.sqrt(res_1p.scale / res_olsf.scale)
# corrct for differences in scale computation
assert_allclose(res_1p.bse, res_olsf.bse * corr_fact, rtol=1e-3)
# regression test, does not verify numbers
# especially why are these smaller than OLS on full sample
# in larger sample, nobs=600, those were close to full OLS
bse1 = np.array([
0.26589869, 0.15224812, 0.38407399, 0.75679949, 0.66084200,
0.54174080, 0.53697607, 0.66006377, 0.38228551, 0.53920485
])
assert_allclose(res_1s.bse, bse1, rtol=1e-7)
def test_combine_subset_regression(self):
# split sample into two, use first sample as prior for second
endog = self.endog
exog = self.exog
nobs = len(endog)
n05 = nobs // 2
np.random.seed(987125)
# shuffle to get random subsamples
shuffle_idx = np.random.permutation(np.arange(nobs))
ys = endog[shuffle_idx]
xs = exog[shuffle_idx]
k = 10
res_ols0 = OLS(ys[:n05], xs[:n05, :k]).fit()
res_ols1 = OLS(ys[n05:], xs[n05:, :k]).fit()
w = res_ols1.scale / res_ols0.scale #1.01
mod_1 = TheilGLS(ys[n05:],
xs[n05:, :k],
r_matrix=np.eye(k),
q_matrix=res_ols0.params,
sigma_prior=w * res_ols0.cov_params())
res_1p = mod_1.fit(cov_type='data-prior')
res_1s = mod_1.fit(cov_type='sandwich')
res_olsf = OLS(ys, xs[:, :k]).fit()
assert_allclose(res_1p.params, res_olsf.params, rtol=1e-9)
corr_fact = 0.96156318 # corrct for differences in scale computation
assert_allclose(res_1p.bse, res_olsf.bse * corr_fact, rtol=1e-3)
# regression test, does not verify numbers
# especially why are these smaller than OLS on full sample
# in larger sample, nobs=600, those were close to full OLS
bse1 = np.array([
0.27609914, 0.15808869, 0.39880789, 0.78583194, 0.68619331,
0.56252314, 0.55757562, 0.68538523, 0.39695081, 0.55988991
])
assert_allclose(res_1s.bse, bse1, rtol=1e-7)
def test_combine_subset_regression(self):
# split sample into two, use first sample as prior for second
endog = self.endog
exog = self.exog
nobs = len(endog)
n05 = nobs // 2
np.random.seed(987125)
# shuffle to get random subsamples
shuffle_idx = np.random.permutation(np.arange(nobs))
ys = endog[shuffle_idx]
xs = exog[shuffle_idx]
k = 10
res_ols0 = OLS(ys[:n05], xs[:n05, :k]).fit()
res_ols1 = OLS(ys[n05:], xs[n05:, :k]).fit()
w = res_ols1.scale / res_ols0.scale #1.01
mod_1 = TheilGLS(ys[n05:], xs[n05:, :k], r_matrix=np.eye(k),
q_matrix=res_ols0.params,
sigma_prior=w * res_ols0.cov_params())
res_1p = mod_1.fit(cov_type='data-prior')
res_1s = mod_1.fit(cov_type='sandwich')
res_olsf = OLS(ys, xs[:, :k]).fit()
assert_allclose(res_1p.params, res_olsf.params, rtol=1e-9)
corr_fact = np.sqrt(res_1p.scale / res_olsf.scale)
# corrct for differences in scale computation
assert_allclose(res_1p.bse, res_olsf.bse * corr_fact, rtol=1e-3)
# regression test, does not verify numbers
# especially why are these smaller than OLS on full sample
# in larger sample, nobs=600, those were close to full OLS
bse1 = np.array([
0.26589869, 0.15224812, 0.38407399, 0.75679949, 0.66084200,
0.54174080, 0.53697607, 0.66006377, 0.38228551, 0.53920485])
assert_allclose(res_1s.bse, bse1, rtol=1e-7)
def test_regression(self):
y = self.endog
x = self.exog
n_groups, k_vars = self.dgp.n_groups, self.dgp.k_vars
R = np.c_[np.zeros((n_groups, k_vars - 1)), np.eye(n_groups)]
r = np.zeros(n_groups)
R = np.c_[np.zeros((n_groups - 1, k_vars)),
np.eye(n_groups - 1) - 1. / n_groups * np.ones(
(n_groups - 1, n_groups - 1))]
r = np.zeros(n_groups - 1)
R[:, k_vars - 1] = -1
lambd = 1 #1e-4
mod = TheilGLS(y, x, r_matrix=R, q_matrix=r, sigma_prior=lambd)
res = mod.fit()
# regression test
params1 = np.array([
0.96518694, 1.06152005, 0.31844136, 3.02747485, 3.25308031,
3.76229199, 1.99795797, 3.9831158, 3.1055317, 1.91599103,
4.5354633, 4.14332517, 3.69462963, 3.79567255, 2.18633118,
2.02848738, 3.74269763, 3.60041509, 3.27734962, 2.47771329,
3.23858674, 4.2973348, 3.98013994, 3.73415254, 2.88870379,
3.91311563, 3.71043309, 1.80506601, 3.78067131, 1.77164485,
3.88247, 3.28328127, 3.1313951, 3.03006754, 3.31012921, 3.08761618,
2.96735903, 1.54005178, 1.27778498, 1.47949121, 4.87184321,
3.03812406, 3.43574332, 2.16983158, 4.45339409, 2.64502381,
4.04767553, 4.42282326, 2.40153298, 3.55409206, 2.71256315,
3.32197196, 3.56054788, 2.58639318, 0.96230275, 1.8382348,
2.30788361, 2.49415769, 0.74777288, 3.04014659, 1.82256153,
4.89165865
])
assert_allclose(res.params, params1)
pen_weight_aicc = mod.select_pen_weight(method='aicc')
pen_weight_gcv = mod.select_pen_weight(method='gcv')
pen_weight_cv = mod.select_pen_weight(method='cv')
pen_weight_bic = mod.select_pen_weight(method='bic')
assert_allclose(pen_weight_gcv, pen_weight_aicc, rtol=0.1)
# regression tests:
assert_allclose(pen_weight_aicc, 2.98779297, rtol=1e-4)
assert_allclose(pen_weight_gcv, 2.69970703, rtol=1e-4)
assert_allclose(pen_weight_bic, 5.76005859, rtol=1e-4)
assert_allclose(pen_weight_cv, 1.3, rtol=1e-4)
def test_regression(self):
y = self.endog
x = self.exog
n_groups, k_vars = self.dgp.n_groups, self.dgp.k_vars
Rg = (np.eye(n_groups - 1) - 1. / n_groups * np.ones(
(n_groups - 1, n_groups - 1)))
R = np.c_[np.zeros((n_groups - 1, k_vars)), Rg]
r = np.zeros(n_groups - 1)
R[:, k_vars - 1] = -1
lambd = 1 #1e-4
mod = TheilGLS(y, x, r_matrix=R, q_matrix=r, sigma_prior=lambd)
res = mod.fit()
# regression test
params1 = np.array([
0.9751655, 1.05215277, 0.37135028, 2.0492626, 2.82062503,
2.82139775, 1.92940468, 2.96942081, 2.86349583, 3.20695368,
4.04516422, 3.04918839, 4.54748808, 3.49026961, 3.15529618,
4.25552932, 2.65471759, 3.62328747, 3.07283053, 3.49485898,
3.42301424, 2.94677593, 2.81549427, 2.24895113, 2.29222784,
2.89194946, 3.17052308, 2.37754241, 3.54358533, 3.79838425,
1.91189071, 1.15976407, 4.05629691, 1.58556827, 4.49941666,
4.08608599, 3.1889269, 2.86203652, 3.06785013, 1.9376162,
2.90657681, 3.71910592, 3.15607617, 3.58464547, 2.15466323,
4.87026717, 2.92909833, 2.64998337, 2.891171, 4.04422964,
3.54616122, 4.12135273, 3.70232028, 3.8314497, 2.2591451,
2.39321422, 3.13064532, 2.1569678, 2.04667506, 3.92064689,
3.66243644, 3.11742725
])
assert_allclose(res.params, params1)
pen_weight_aicc = mod.select_pen_weight(method='aicc')
pen_weight_gcv = mod.select_pen_weight(method='gcv')
pen_weight_cv = mod.select_pen_weight(method='cv')
pen_weight_bic = mod.select_pen_weight(method='bic')
assert_allclose(pen_weight_gcv, pen_weight_aicc, rtol=0.1)
# regression tests:
assert_allclose(pen_weight_aicc, 4.77333984, rtol=1e-4)
assert_allclose(pen_weight_gcv, 4.45546875, rtol=1e-4)
assert_allclose(pen_weight_bic, 9.35957031, rtol=1e-4)
assert_allclose(pen_weight_cv, 1.99277344, rtol=1e-4)
def test_regression(self):
y = self.endog
x = self.exog
n_groups, k_vars = self.dgp.n_groups, self.dgp.k_vars
R = np.c_[np.zeros((n_groups, k_vars-1)), np.eye(n_groups)]
r = np.zeros(n_groups)
R = np.c_[np.zeros((n_groups-1, k_vars)),
np.eye(n_groups-1)-1./n_groups * np.ones((n_groups-1, n_groups-1))]
r = np.zeros(n_groups-1)
R[:, k_vars-1] = -1
lambd = 1 #1e-4
mod = TheilGLS(y, x, r_matrix=R, q_matrix=r, sigma_prior=lambd)
res = mod.fit()
# regression test
params1 = np.array([
0.96518694, 1.06152005, 0.31844136, 3.02747485, 3.25308031,
3.76229199, 1.99795797, 3.9831158 , 3.1055317 , 1.91599103,
4.5354633 , 4.14332517, 3.69462963, 3.79567255, 2.18633118,
2.02848738, 3.74269763, 3.60041509, 3.27734962, 2.47771329,
3.23858674, 4.2973348 , 3.98013994, 3.73415254, 2.88870379,
3.91311563, 3.71043309, 1.80506601, 3.78067131, 1.77164485,
3.88247 , 3.28328127, 3.1313951 , 3.03006754, 3.31012921,
3.08761618, 2.96735903, 1.54005178, 1.27778498, 1.47949121,
4.87184321, 3.03812406, 3.43574332, 2.16983158, 4.45339409,
2.64502381, 4.04767553, 4.42282326, 2.40153298, 3.55409206,
2.71256315, 3.32197196, 3.56054788, 2.58639318, 0.96230275,
1.8382348 , 2.30788361, 2.49415769, 0.74777288, 3.04014659,
1.82256153, 4.89165865])
assert_allclose(res.params, params1)
pen_weight_aicc = mod.select_pen_weight(method='aicc')
pen_weight_gcv = mod.select_pen_weight(method='gcv')
pen_weight_cv = mod.select_pen_weight(method='cv')
pen_weight_bic = mod.select_pen_weight(method='bic')
assert_allclose(pen_weight_gcv, pen_weight_aicc, rtol=0.1)
# regression tests:
assert_allclose(pen_weight_aicc, 2.98779297, rtol=1e-4)
assert_allclose(pen_weight_gcv, 2.69970703, rtol=1e-4)
assert_allclose(pen_weight_bic, 5.76005859, rtol=1e-4)
assert_allclose(pen_weight_cv, 1.3, rtol=1e-4)
def test_regression(self):
y = self.endog
x = self.exog
n_groups, k_vars = self.dgp.n_groups, self.dgp.k_vars
Rg = (np.eye(n_groups-1) - 1. / n_groups *
np.ones((n_groups - 1, n_groups-1)))
R = np.c_[np.zeros((n_groups - 1, k_vars)), Rg]
r = np.zeros(n_groups - 1)
R[:, k_vars-1] = -1
lambd = 1 #1e-4
mod = TheilGLS(y, x, r_matrix=R, q_matrix=r, sigma_prior=lambd)
res = mod.fit()
# regression test
params1 = np.array([
0.9751655 , 1.05215277, 0.37135028, 2.0492626 , 2.82062503,
2.82139775, 1.92940468, 2.96942081, 2.86349583, 3.20695368,
4.04516422, 3.04918839, 4.54748808, 3.49026961, 3.15529618,
4.25552932, 2.65471759, 3.62328747, 3.07283053, 3.49485898,
3.42301424, 2.94677593, 2.81549427, 2.24895113, 2.29222784,
2.89194946, 3.17052308, 2.37754241, 3.54358533, 3.79838425,
1.91189071, 1.15976407, 4.05629691, 1.58556827, 4.49941666,
4.08608599, 3.1889269 , 2.86203652, 3.06785013, 1.9376162 ,
2.90657681, 3.71910592, 3.15607617, 3.58464547, 2.15466323,
4.87026717, 2.92909833, 2.64998337, 2.891171 , 4.04422964,
3.54616122, 4.12135273, 3.70232028, 3.8314497 , 2.2591451 ,
2.39321422, 3.13064532, 2.1569678 , 2.04667506, 3.92064689,
3.66243644, 3.11742725])
assert_allclose(res.params, params1)
pen_weight_aicc = mod.select_pen_weight(method='aicc')
pen_weight_gcv = mod.select_pen_weight(method='gcv')
pen_weight_cv = mod.select_pen_weight(method='cv')
pen_weight_bic = mod.select_pen_weight(method='bic')
assert_allclose(pen_weight_gcv, pen_weight_aicc, rtol=0.1)
# regression tests:
assert_allclose(pen_weight_aicc, 4.77333984, rtol=1e-4)
assert_allclose(pen_weight_gcv, 4.45546875, rtol=1e-4)
assert_allclose(pen_weight_bic, 9.35957031, rtol=1e-4)
assert_allclose(pen_weight_cv, 1.99277344, rtol=1e-4)
def setup_class(cls):
y, x = cls.get_sample()
mod1 = TheilGLS(y, x, sigma_prior=[0, 0, 1., 1.])
cls.res1 = mod1.fit(0)
cls.res2 = OLS(y, x).fit()
def setup_class(cls):
y, x = cls.get_sample()
mod1 = TheilGLS(y, x, sigma_prior=[0, 0, 1., 1.])
cls.res1 = mod1.fit(0)
cls.res2 = OLS(y, x).fit()
