def test_standardize1():
np.random.seed(123)
x = 1 + np.random.randn(5, 4)
transf = StandardizeTransform(x)
xs1 = transf(x)
assert_allclose(transf.mean, x.mean(0), rtol=1e-13)
assert_allclose(transf.scale, x.std(0, ddof=1), rtol=1e-13)
xs2 = stats.zscore(x, ddof=1)
assert_allclose(xs1, xs2, rtol=1e-13, atol=1e-20)
# check we use stored transformation
xs4 = transf(2 * x)
assert_allclose(xs4, (2 * x - transf.mean) / transf.scale,
rtol=1e-13,
atol=1e-20)
# affine transform doesn't change standardized
x2 = 2 * x + np.random.randn(4)
transf2 = StandardizeTransform(x2)
xs3 = transf2(x2)
assert_allclose(xs3, xs1, rtol=1e-13, atol=1e-20)
# check constant
x5 = np.column_stack((np.ones(x.shape[0]), x))
transf5 = StandardizeTransform(x5)
xs5 = transf5(x5)
assert_equal(transf5.const_idx, 0)
assert_equal(xs5[:, 0], np.ones(x.shape[0]))
assert_allclose(xs5[:, 1:], xs1, rtol=1e-13, atol=1e-20)
def test_standardize_ols():
np.random.seed(123)
nobs = 20
x = 1 + np.random.randn(nobs, 4)
exog = np.column_stack((np.ones(nobs), x))
endog = exog.sum(1) + np.random.randn(nobs)
res2 = OLS(endog, exog).fit()
transf = StandardizeTransform(exog)
exog_st = transf(exog)
res1 = OLS(endog, exog_st).fit()
params = transf.transform_params(res1.params)
assert_allclose(params, res2.params, rtol=1e-13)
def setupClass(cls):
data = sm.datasets.spector.load()
data.exog = sm.add_constant(data.exog, prepend=False)
res2 = Spector()
res2.probit()
cls.res2 = res2
# fmin_cg fails to converge on some machines - reparameterize
from statsmodels.tools.transform_model import StandardizeTransform
transf = StandardizeTransform(data.exog)
exog_st = transf(data.exog)
res1_st = Probit(data.endog, exog_st).fit(method="cg",
disp=0, maxiter=500, gtol=1e-08)
start_params = transf.transform_params(res1_st.params)
assert_allclose(start_params, res2.params, rtol=1e-5, atol=1e-6)
cls.res1 = Probit(data.endog, data.exog).fit(start_params=start_params,
method="cg",
maxiter=500, gtol=1e-08,
disp=0)
assert_array_less(cls.res1.mle_retvals['fcalls'], 70)