def test_pls_load_plot_without_labels():
pls = nipals.PLS(oliveoil_missing_y.iloc[:, :5],
oliveoil_missing_y.iloc[:, 5:])
assert pls.fit(ncomp=2)
plt = pls.loadingsplot(labels=False)
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_cv_pls():
pls = nipals.PLS(oliveoil.iloc[:, :5], oliveoil.iloc[:, 5:])
assert pls.fit(ncomp=2, cv=True)
assert pls.fit(ncomp=2, cv=4)
np.testing.assert_almost_equal(pls.Q2, [0.2684853, 0.0644187])
np.testing.assert_almost_equal(pls.R2X, [0.5826444, 0.2367458])
np.testing.assert_almost_equal(pls.R2Y, [0.4326835, 0.0856207])
def test_pls_load_plot_with_markers():
pls = nipals.PLS(oliveoil_missing_y.iloc[:, :5],
oliveoil_missing_y.iloc[:, 5:])
assert pls.fit(ncomp=2)
plt = pls.loadingsplot(weightmarkers=['s', 'o', 'v', '^', '+', '3'])
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_pls_nocenter_and_scale():
pls = nipals.PLS(yarn.iloc[:, :268] - yarn.iloc[:, :268].mean(),
yarn.iloc[:, 268])
assert pls.fit(ncomp=6, scale=False, center=False)
np.testing.assert_almost_equal(pls.scores.values, yarn_scores)
np.testing.assert_almost_equal(pls.loadings.values, yarn_loadings)
np.testing.assert_almost_equal(pls.weights.values, yarn_weights)
def test_pls_score_plot():
pls = nipals.PLS(oliveoil_missing_y.iloc[:, :5],
oliveoil_missing_y.iloc[:, 5:])
assert pls.fit(ncomp=2)
plt = pls.plot()
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_pls_zerovar_y():
tmpy = oliveoil.iloc[:, 5:].copy()
tmpy.loc[:, 'extra'] = 2.0
pls = nipals.PLS(oliveoil.iloc[:, :5], tmpy)
with pytest.raises(ValueError) as e_info:
pls.fit(2)
assert e_info.match("zero variance in column.*extra")
assert pls.fit(2, dropzerovar=True)
def test_pls_optionvariations(caplog):
pls = nipals.PLS(yarn.iloc[:, :268], yarn.iloc[:, 268])
assert pls.fit()
assert pls.fit(ncomp=500, startcol=0)
assert caplog.record_tuples == [
('root', logging.WARNING,
'ncomp is larger than the max dimension of the x matrix.\n'
'fit will only return 28 components'),
]
with pytest.raises(RuntimeError):
pls.fit(maxiter=1)
def test_pls_missing_x():
tmpx = yarn.iloc[:, :268]
csel = [171, 107, 222, 150, 76, 205, 63, 19, 121, 183]
rsel = [23, 0, 3, 22, 15, 21, 19, 7, 19, 5]
for r, c in zip(rsel, csel):
tmpx.iloc[r, c] = np.nan
pls = nipals.PLS(tmpx, yarn.iloc[:, 268])
assert pls.fit(ncomp=3)
np.testing.assert_almost_equal(pls.scores.values, yarn_missing_scores, 5)
np.testing.assert_almost_equal(pls.loadings.values, yarn_missing_loadings)
np.testing.assert_almost_equal(pls.weights.values, yarn_missing_weights)
def test_inf_values_in_dfs(caplog):
olive_inf = oliveoil.replace([18.7, 75.1], np.inf)
nipals.Nipals(olive_inf)
nipals.PLS(olive_inf.iloc[:, :5], olive_inf.iloc[:, 5:])
assert caplog.record_tuples == [
('root', logging.WARNING,
'Data contained infinite values, converting to missing values'),
('root', logging.WARNING,
'X data contained infinite values, converting to missing values'),
('root', logging.WARNING,
'Y data contained infinite values, converting to missing values'),
]
def test_dmod_pls():
pls = nipals.PLS(oliveoil.iloc[:, :5], oliveoil.iloc[:, 5:])
assert pls.fit(ncomp=2)
np.testing.assert_almost_equal(pls.dModX(), [
0.8561775, 0.6808639, 1.2475466, 1.9522224, 0.1694258, 0.6151471,
0.8651022, 1.1295028, 1.2212977, 1.2945167, 0.8681262, 1.141947,
0.7637504, 0.3900809, 0.7382779, 0.7063927
])
np.testing.assert_almost_equal(pls.dModY(), [
0.5728738, 2.0598601, 1.2420525, 0.5257193, 1.4290988, 0.7752674,
1.0029673, 0.8943648, 1.0509669, 0.8511583, 0.7153933, 1.0285289,
0.8238886, 0.4616424, 0.5664762, 0.7409856
])
def test_overviewplotplot_pls():
pls = nipals.PLS(oliveoil.iloc[:, :5], oliveoil.iloc[:, 5:])
assert pls.fit(ncomp=2)
plt = pls.overviewplot()
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_pls_missing_y():
pls = nipals.PLS(oliveoil_missing_y.iloc[:, :5],
oliveoil_missing_y.iloc[:, 5:])
assert pls.fit(ncomp=2)
np.testing.assert_almost_equal(pls.scores.values,
oliveoil_missing_y_scores * [-1, 1], 3)
def test_pls_nondf():
pls = nipals.PLS(yarn.iloc[:, :268].values, yarn.iloc[:, 268].values)
assert pls.fit(ncomp=6, scale=False)
np.testing.assert_almost_equal(pls.scores.values, yarn_scores)
np.testing.assert_almost_equal(pls.loadings.values, yarn_loadings)
np.testing.assert_almost_equal(pls.weights.values, yarn_weights)