def test_numpy_vector_input(self):
x = randn(3)
detrends = detrend_mean, detrend_none, detrend_linear
scale_types = 'normalize', 'none', 'unbiased', 'biased', None
maxlags = 2, None, 100
args = cartprod(maxlags, detrends, scale_types)
for ml, dt, st in args:
y = randn(randint(2, 3))
if ml > max(x.shape + y.shape):
self.assertRaises(AssertionError, xcorr, x, detrend=dt,
scale_type=st, maxlags=ml)
self.assertRaises(AssertionError, xcorr, x, y, detrend=dt,
scale_type=st, maxlags=ml)
self.assertRaises(AssertionError, xcorr, x, y=None, detrend=dt,
scale_type=st, maxlags=ml)
else:
xcnn = xcorr(x, detrend=dt, scale_type=st, maxlags=ml)
xcnn2 = xcorr(x, y, detrend=dt, scale_type=st, maxlags=ml)
xcnn3 = xcorr(x, y=None, detrend=dt, scale_type=st, maxlags=ml)
self.assertIsInstance(xcnn, np.ndarray)
self.assertIsInstance(xcnn2, np.ndarray)
self.assertIsInstance(xcnn3, np.ndarray)
def test_numpy_matrix_input(self):
x = randn(10, 15)
detrends = detrend_mean, detrend_none, detrend_linear
scale_types = 'normalize', None, 'unbiased', 'biased'
maxlags = 8, None, 100
args = cartprod(maxlags, detrends, scale_types)
for ml, dt, st in args:
if ml > x.shape[0]:
self.assertRaises(AssertionError, xcorr, x, detrend=dt,
scale_type=st, maxlags=ml)
else:
xcnn = xcorr(x, detrend=dt, scale_type=st, maxlags=ml)
self.assertIsInstance(xcnn, np.ndarray)
def test_matrix_vs_numpy_correlate(self):
m, n = 5, 3
x = randn(m, n)
rng = xrange(n)
detrends = detrend_mean, detrend_none, detrend_linear
scale_types = 'normalize', None, 'unbiased', 'biased'
maxlags = None, 2, 20
args = cartprod(maxlags, detrends, scale_types)
xc_np = np.zeros((2 * m - 1, n ** 2))
cart = cartesian((np.arange(n), np.arange(n)))
cols = map(tuple, cart)
for ml, dt, st in args:
xc_np.fill(0)
mml = ml if ml is not None else m
xc_span = DataFrame(xc_np.copy(), columns=cols,
index=Int64Index(np.r_[1 - m:m]))
for ci, (i, j) in enumerate(cartprod(rng, rng)):
xi, xj = dt(x[:, i]), dt(x[:, j])
xc_np[:, ci] = np.correlate(xi, xj, mode='full')
if st == 'normalize':
cxx0 = sum(abs(xi) ** 2)
cyy0 = sum(abs(xj) ** 2)
xc_np[:, ci] /= np.sqrt(cxx0 * cyy0)
elif st == 'biased':
xc_np[:, ci] /= m
elif st == 'unbiased':
xc_np[:, ci] /= m - np.abs(np.r_[1 - m:m])
xc_span[i, j] = xcorr(xi, xj, detrend=dt, scale_type=st)
if ml > m:
self.assertRaises(AssertionError, xcorr, x, detrend=dt,
scale_type=st, maxlags=ml)
else:
xcc = xcorr(x, detrend=dt, scale_type=st, maxlags=ml)
assert_allclose(xc_np, xc_span)
dd = DataFrame(xc_np, index=np.r_[1 - m:m])
# funky pandas df indexing must subtract one because endpoints
# are inclusive
assert_allclose(dd.ix[1 - mml:mml - 1].values, xcc)
self.assertIsInstance(xcc, np.ndarray)
if ml > m:
self.assertRaises(AssertionError, xcorr, DataFrame(x),
detrend=dt, scale_type=st, maxlags=ml)
else:
xccdf = xcorr(DataFrame(x), detrend=dt, scale_type=st,
maxlags=ml)
assert_allclose(xccdf, xcc)
self.assertIsInstance(xccdf, DataFrame)
if ml > m:
self.assertRaises(AssertionError, xcorr, Series(x[:, 0]),
detrend=dt, scale_type=st, maxlags=ml)
self.assertRaises(AssertionError, xcorr, x[:, 0],
detrend=dt, scale_type=st, maxlags=ml)
else:
xc_s = xcorr(Series(x[:, 0]), detrend=dt, scale_type=st,
maxlags=ml)
xc_s2 = xcorr(Series(x[:, 0]), y=None, detrend=dt,
scale_type=st, maxlags=ml)
assert_allclose(xc_s, xc_s2)
if ml > m:
self.assertRaises(AssertionError, xcorr, Series(x[:, 0]),
Series(x[:, 1]), detrend=dt, scale_type=st,
maxlags=ml)
self.assertRaises(AssertionError, xcorr, x[:, 0], x[:, 1],
detrend=dt, scale_type=st, maxlags=ml)
else:
xc_s2 = xcorr(Series(x[:, 0]), Series(x[:, 1]), detrend=dt,
scale_type=st, maxlags=ml)
xc_s2_np = xcorr(x[:, 0], x[:, 1], detrend=dt, scale_type=st,
maxlags=ml)
assert_allclose(xc_s2, xc_s2_np)