def test_shuffle_vs_alpha(self):
print("Checking that bootstrap conf intervals equal analytical ones "
"for conf_lev = %.2f, conf_samples = %d" % (self.conf_lev,
self.conf_samples))
for measure in ['reg', 'par_corr', 'cmi_gauss']:
tmp = te._get_estimate(array=self.array,
measure=measure, xyz=self.xyz,
confidence='analytic',
conf_samples=self.conf_samples,
conf_lev=self.conf_lev,
measure_params=self.measure_params,
verbosity=0)
expected = tmp['conf_lower'], tmp['conf_upper']
tmp = te._get_estimate(array=self.array,
measure=measure, xyz=self.xyz,
confidence='bootstrap',
conf_samples=self.conf_samples,
conf_lev=self.conf_lev,
measure_params=self.measure_params,
verbosity=0)
res = tmp['conf_lower'], tmp['conf_upper']
print("bootstrap interval for %s = (%.3f, %.3f) -- "
"analytic interval = (%.3f, %.3f)" % (
measure, res[0], res[1], expected[0], expected[1]))
numpy.testing.assert_allclose(res, expected, rtol=self.rtol)
def test_shuffle_vs_alpha(self):
print("Checking that bootstrap conf intervals equal analytical ones "
"for conf_lev = %.2f, conf_samples = %d" %
(self.conf_lev, self.conf_samples))
for measure in ['reg', 'par_corr', 'cmi_gauss']:
tmp = te._get_estimate(array=self.array,
measure=measure,
xyz=self.xyz,
confidence='analytic',
conf_samples=self.conf_samples,
conf_lev=self.conf_lev,
measure_params=self.measure_params,
verbosity=0)
expected = tmp['conf_lower'], tmp['conf_upper']
tmp = te._get_estimate(array=self.array,
measure=measure,
xyz=self.xyz,
confidence='bootstrap',
conf_samples=self.conf_samples,
conf_lev=self.conf_lev,
measure_params=self.measure_params,
verbosity=0)
res = tmp['conf_lower'], tmp['conf_upper']
print("bootstrap interval for %s = (%.3f, %.3f) -- "
"analytic interval = (%.3f, %.3f)" %
(measure, res[0], res[1], expected[0], expected[1]))
numpy.testing.assert_allclose(res, expected, rtol=self.rtol)
def test_shuffle_vs_alpha(self):
print("Checking that shuffle sig_thres equals analytical thres "
"for sig_lev = %.2f, sig_samples = %d" % (self.sig_lev,
self.sig_samples))
numpy.random.seed(42)
# array = numpy.random.randn(2, self.T)
links_coeffs = {0: [],
1: [((0, 0), .06)], }
data, links = pp.var_process(links_coeffs, T=self.T,
use='inno_cov', verbosity=verbosity)
array = data.T
xyz = numpy.array([0, 1])
for measure in ['par_corr', 'reg']:
expected = te._get_estimate(array=array, measure=measure, xyz=xyz,
significance='analytic',
sig_samples=self.sig_samples,
sig_lev=self.sig_lev,
measure_params=self.measure_params,
verbosity=0)['sig_thres']
res = te._get_estimate(array=array, measure=measure, xyz=xyz,
significance='full_shuffle',
sig_samples=self.sig_samples,
sig_lev=self.sig_lev,
measure_params=self.measure_params,
verbosity=0)['sig_thres']
print("shuffle %.2f sig thres for %s = %.4f (analytic = %.4f)"
% (self.sig_lev, measure, res, expected))
numpy.testing.assert_allclose(res, expected, rtol=self.rtol)
def test_shuffle_vs_alpha(self):
print("Checking that shuffle sig_thres equals analytical thres "
"for sig_lev = %.2f, sig_samples = %d" %
(self.sig_lev, self.sig_samples))
numpy.random.seed(42)
# array = numpy.random.randn(2, self.T)
links_coeffs = {
0: [],
1: [((0, 0), .06)],
}
data, links = pp.var_process(links_coeffs,
T=self.T,
use='inno_cov',
verbosity=verbosity)
array = data.T
xyz = numpy.array([0, 1])
for measure in ['par_corr', 'reg']:
expected = te._get_estimate(array=array,
measure=measure,
xyz=xyz,
significance='analytic',
sig_samples=self.sig_samples,
sig_lev=self.sig_lev,
measure_params=self.measure_params,
verbosity=0)['sig_thres']
res = te._get_estimate(array=array,
measure=measure,
xyz=xyz,
significance='full_shuffle',
sig_samples=self.sig_samples,
sig_lev=self.sig_lev,
measure_params=self.measure_params,
verbosity=0)['sig_thres']
print("shuffle %.2f sig thres for %s = %.4f (analytic = %.4f)" %
(self.sig_lev, measure, res, expected))
numpy.testing.assert_allclose(res, expected, rtol=self.rtol)
