def test_windowed_pass_nanmedian():
raw = [1, 2, 3, np.nan, np.nan, 4, 5, 6, 7, np.nan]
data = np.array(raw, dtype='float')
assert nanmedian(data) == approx(4.0)
res = windowed_pass(data, 5, value=nanmedian)
assert np.all(np.isnan(res[:4]))
assert res[4] == approx(2.0)
assert res[5] == approx(3.0)
assert res[6] == approx(4.0)
assert res[7] == approx(5.0)
assert res[8] == approx(5.5)
res2 = windowed_pass(data, 2, value=nanmedian)
assert np.isnan(res2[0])
assert res2[1] == approx(1.5)
assert res2[2] == approx(2.5)
assert res2[3] == approx(3.0)
assert np.isnan(res2[4])
assert res2[5] == approx(4.0)
assert res2[6] == approx(4.5)
assert res2[7] == approx(5.5)
assert res2[8] == approx(6.5)
assert res2[9] == approx(7.0)
def test_windowed_pass_ols_r_squared():
"""
Returns a 2-column array the same size
as `data` with the OLS slope values
in the first column and the r^2 values
in the second.
"""
data = np.array([[0, 1], [1, 2], [2, 3], [3, 4], [2, 3], [1, 2], [0, 1]],
dtype='float')
raw = ols_r_squared(data)
assert raw[0] == approx(1.47368421)
assert raw[1] == approx(0.63157895)
res = windowed_pass(data, 3, value=ols_r_squared)
assert np.isnan(res[0, 0])
assert np.isnan(res[0, 1])
assert np.isnan(res[1, 0])
assert np.isnan(res[1, 1])
assert res[2, 0] == approx(1.6)
assert res[2, 1] == approx(0.4)
assert res[3, 0] == approx(1.42857143)
assert res[3, 1] == approx(0.78571429)
assert res[4, 0] == approx(1.41176471)
assert res[4, 1] == approx(0.82352941)
assert res[5, 0] == approx(1.42857143)
assert res[5, 1] == approx(0.78571429)
assert res[6, 0] == approx(1.6)
assert res[6, 1] == approx(0.4)
def test_windowed_pass_ols_wrapped():
"""
Returns a 2-column array the same size
as `data` with the OLS slope values
duplicated in both columns.
"""
data = np.array(
[
[0, 1], # np.nan
[1, 2], # np.nan
[2, 3], # 1.40000000
[3, 4], # 1.42857143
[2, 3], # 1.41176471
[1, 2], # 1.42857143
[0, 1] # 1.60000000
],
dtype='float')
assert ols_wrap(data) == approx(1.4736842105263157)
res = windowed_pass(data, 3, value=ols_wrap)
# Two values are returned here as we haven't
# added support for multiple args yet
assert np.isnan(res[0, 0])
assert np.isnan(res[1, 0])
assert res[2, 0] == approx(1.6)
assert res[3, 0] == approx(1.42857143)
assert res[4, 0] == approx(1.41176471)
assert res[5, 0] == approx(1.42857143)
assert res[6, 0] == approx(1.6)
def test_windowed_pass_sum():
data = np.array(range(100), dtype='float')
assert nsum(data) == approx(4950)
res = windowed_pass(data, 3, value=nsum)
assert np.isnan(res[0])
assert np.isnan(res[1])
assert res[2] == approx(3.0)
assert res[3] == approx(6.0)
assert res[4] == approx(9.0)
assert res[-1] == approx(294.0)
res5 = windowed_pass(data, 5, value=nsum)
assert np.all(np.isnan(res5[:4]))
assert res5[5] == approx(15.0)
assert res5[-1] == approx(485.0)
def test_windowed_pass_mean_size_10():
data = np.array(range(100), dtype='float')**2
res = windowed_pass(data, 10, value=mean)
assert np.all(np.isnan(res[:9]))
assert res[9] == approx(28.5)
assert res[10] == approx(38.5)
assert res[11] == approx(50.5)
assert res[-1] == approx(8938.5)
def test_windowed_pass_basic_sanity():
assert std([1, 2, 3]) == approx(0.816496580)
data = np.array(range(100), dtype='float')**2
res = windowed_pass(data, 5, value=std)
assert np.isnan(res[0])
assert np.isnan(res[1])
assert np.isnan(res[2])
assert np.isnan(res[3])
assert res[4] == approx(5.89915248)
assert res[5] == approx(8.64869932)
assert res[-1] == approx(274.36253389)
raw = windowed_pass(data, 5)
assert np.isnan(raw[3])
assert raw[4] == approx(0.0)
assert raw[50] == approx(2116.0)
def test_windowed_pass_mean_size_5():
assert mean(np.array([1, 2, 3])) == approx(2.0)
data = np.array(range(100), dtype='float')**2
res = windowed_pass(data, 5, value=mean)
assert np.isnan(res[0])
assert np.isnan(res[1])
assert np.isnan(res[3])
assert res[4] == approx(6.0)
assert res[5] == approx(11.0)
assert res[8] == approx(38.0)
assert res[-1] == approx(9411.0)