def test_run_pca_without_na():
nip = nipals.Nipals(testdata_full)
assert nip.fit()
np.testing.assert_almost_equal(list(nip.eig), [
5.020518433605382, 1.879323465996815, 1.1081766447275905,
0.17225187199265019, 0.06936702860594454
])
def test_cv_pca_with_fliped_axis():
"""Checks that the Q2 is approx the same when some cvs have flipped axis in some PC"""
nip = nipals.Nipals(testdata_class)
assert nip.fit(ncomp=2, cv=True, startcol=2)
Q2_sc2 = nip.Q2.values
assert nip.fit(ncomp=2, cv=True)
np.testing.assert_almost_equal(nip.Q2, Q2_sc2, 3)
def test_predict_from_pca():
nip = nipals.Nipals(testdata)
nip.fit(ncomp=2)
assert nip.predict(
pd.DataFrame([[63, 70, 98, 110, 124], [51, 82, 102, 110, 108]]))
np.testing.assert_almost_equal(
nip.pred.values, [[-1.4465766, 0.4500705], [-1.6229739, -0.340578]])
def test_run_pca_with_prescaled_data():
scaled = pd.DataFrame(testdata_full)
scaled = (scaled - scaled.mean()) / scaled.std(ddof=1)
nip = nipals.Nipals(scaled)
assert nip.fit(center=False, scale=False, ncomp=2)
np.testing.assert_almost_equal(list(nip.eig),
[5.020518433605, 1.879323465996])
def test_dmodx_pca():
nip = nipals.Nipals(testdata_class)
assert nip.fit(ncomp=2)
np.testing.assert_almost_equal(nip.dModX(), [
0.5364634, 0.3480396, 0.5702191, 1.5525099, 1.5752593, 0.3275259,
1.1255771
])
def test_predict_from_pca_with_sweep():
nip = nipals.Nipals(testdata)
nip.fit(ncomp=2, eigsweep=True)
assert nip.predict(
pd.DataFrame([[63, 70, 98, 110, 124], [51, 82, 102, 110, 108]]))
np.testing.assert_almost_equal(
nip.pred.values, [[-0.2966596, 0.2201614], [-0.3328347, -0.1666008]])
def test_run_pca_with_precentered_data():
centered = pd.DataFrame(testdata_full)
centered = centered - centered.mean()
nip = nipals.Nipals(centered)
assert nip.fit(center=False, ncomp=2)
np.testing.assert_almost_equal(list(nip.eig),
[5.020518433605, 1.879323465996])
def test_pca_zerovar():
tmpx = pd.DataFrame(testdata)
tmpx.loc[:, 'extra'] = 2.0
pca = nipals.Nipals(tmpx)
with pytest.raises(ValueError) as e_info:
pca.fit(2)
assert e_info.match("zero variance in column.*extra")
assert pca.fit(2, dropzerovar=True)
def test_pred_plot():
nip = nipals.Nipals(testdata_class)
nip.fit(ncomp=2)
nip.predict(pd.DataFrame([[63, 70, 98, 110, 124], [51, 82, 102, 110,
108]]))
plt = nip.plot(classlevels=['C1', 'C2'], plotpred=True)
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_plot_classoptions():
nip = nipals.Nipals(testdata_class)
nip.fit(ncomp=2)
plt = nip.plot(classlevels=['C1', 'C2'],
markers=['s', 'o'],
classcolors=['red', 'black'])
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_call_with_too_large_ncomp(caplog):
nip = nipals.Nipals(testdata)
assert nip.fit(ncomp=10)
assert caplog.record_tuples == [
('root', logging.WARNING,
'ncomp is larger than the max dimension of the x matrix.\n'
'fit will only return 5 components'),
]
def test_run_pca_check_scores():
nip = nipals.Nipals(testdata)
assert nip.fit(ncomp=2)
np.testing.assert_almost_equal(
nip.scores.values,
[[-2.72332498, 0.25021637], [-1.85369271, 0.34827344],
[-0.98854112, 0.64658032], [-0.16429534, -1.41375825],
[0.62686116, -0.92008715], [1.7370958, -0.41837244],
[2.91721621, 0.746181]])
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_run_pca_check_scores_with_sweep():
nip = nipals.Nipals(testdata)
assert nip.fit(ncomp=2, eigsweep=True, startcol=1)
np.testing.assert_almost_equal(nip.scores.values, [
[-0.5585132, 0.1224190],
[-0.3801627, 0.1703718],
[-0.2026926, 0.3163937],
[-0.0337608, -0.6915786],
[0.1285239, -0.4501362],
[0.3562110, -0.2048250],
[0.5982563, 0.3647261],
], 3)
def test_pred_plot_labels():
nip = nipals.Nipals(testdata_class)
nip.fit(ncomp=2)
nip.predict(
pd.DataFrame([[63, 70, 98, 110, 124, 0, 1, 0],
[51, 82, 102, 110, 108, 1, 0, 0]],
index=['a', 'b']).set_index([5, 6, 7], append=True))
plt = nip.plot(classlevels=['C1', 'C2'],
plotpred=True,
labels='C1',
predlevels=[5, 6],
predlabels=0)
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_predict_from_pca_precentered():
centered = pd.DataFrame(testdata_full)
origmean = centered.mean()
centered = centered - origmean
nip = nipals.Nipals(centered)
assert nip.fit(center=False, ncomp=2)
# assert nip.fit(center=True, ncomp=2)
np.testing.assert_almost_equal(list(nip.eig),
[5.020518433605, 1.879323465996])
predtest = pd.DataFrame([[63, 70, 98, 110, 124], [51, 82, 102, 110, 108]])
predtest = predtest - origmean
assert nip.predict(predtest)
np.testing.assert_almost_equal(
nip.pred.values, [[-1.2749312, 0.7863909], [-1.5595433, 0.0705577]])
def test_predict_from_pca_prescaled():
scaled = pd.DataFrame(testdata_full)
origstd = scaled.std(ddof=1)
scaled = scaled / origstd
nip = nipals.Nipals(scaled)
assert nip.fit(scale=False, ncomp=2)
# assert nip.fit(center=True, ncomp=2)
np.testing.assert_almost_equal(list(nip.eig),
[5.020518433605, 1.879323465996])
predtest = pd.DataFrame([[63, 70, 98, 110, 124], [51, 82, 102, 110, 108]])
predtest = predtest / origstd
assert nip.predict(predtest)
np.testing.assert_almost_equal(
nip.pred.values, [[-1.2749312, 0.7863909], [-1.5595433, 0.0705577]], 3)
def test_pred_plot_options():
nip = nipals.Nipals(testdata_class)
nip.fit(ncomp=2)
nip.predict(
pd.DataFrame([[63, 70, 98, 110, 124, 0, 1, 0],
[51, 82, 102, 110, 108, 1, 0,
0]]).set_index([5, 6, 7], append=True))
with pytest.raises(KeyError):
plt = nip.plot(classlevels=['C1', 'C2'],
plotpred=True,
predlevels=[5, 7])
plt = nip.plot(classlevels=['C1', 'C2'], plotpred=True, predlevels=[5, 6])
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_run_pca():
nip = nipals.Nipals(testdata)
# Set startcol to get same results as R nipals:nipals (rounding errors give slightly different rotation otherwise)
assert nip.fit(startcol=1)
np.testing.assert_almost_equal(list(nip.eig), [
4.876206673116805, 2.044242214135278, 1.072810327696312,
0.23701331933712338, 0.14325051166784325
])
# Also run without startcol set to make sure that it works as well. But compare to self data
assert nip.fit()
np.testing.assert_almost_equal(list(nip.eig), [
4.876216689582536, 2.044275687396918, 1.072805497059184,
0.23696073749622645, 0.14327789003413574
])
def test_run_pca_with_set_ncomp():
nip = nipals.Nipals(testdata_full)
assert nip.fit(ncomp=2)
np.testing.assert_almost_equal(list(nip.eig),
[5.020518433605, 1.879323465996])
def test_fail_from_maxiter():
nip = nipals.Nipals(testdata_full)
with pytest.raises(RuntimeError):
nip.fit(maxiter=1)
def test_plot():
nip = nipals.Nipals(testdata)
nip.fit(ncomp=2)
plt = nip.plot()
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_plot_classes():
nip = nipals.Nipals(testdata_class)
nip.fit(ncomp=2)
plt = nip.plot(classlevels=['C1', 'C2'])
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_init_from_data():
# It should be possible to init a nipals class from something that
# can be made into a Pandas DataFrame
assert nipals.Nipals(testdata).__class__ == nipals.Nipals
def test_overviewplotplot_pca():
nip = nipals.Nipals(testdata_class)
assert nip.fit(ncomp=2)
plt = nip.overviewplot()
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_dmodx_plot_pca():
nip = nipals.Nipals(testdata_class)
assert nip.fit(ncomp=2)
plt = nip.dModXPlot()
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_loadings_plot():
nip = nipals.Nipals(testdata_class)
nip.fit(ncomp=2)
plt = nip.loadingsplot()
assert isinstance(plt, matplotlib.figure.Figure)
return plt
def test_init_from_df():
# It should be possible to init a nipals class from a Pandas DataFrame
assert nipals.Nipals(pd.DataFrame(testdata)).__class__ == nipals.Nipals
def test_cv_pca():
nip = nipals.Nipals(testdata_class)
assert nip.fit(ncomp=2, cv=True)
assert nip.fit(ncomp=2, cv=4)
np.testing.assert_almost_equal(nip.R2, [0.8112004, 0.144992])
np.testing.assert_almost_equal(nip.Q2, [0.577648, 0.1909286])
