def test_vifs():
df = pd.DataFrame(np.random.randn(100, 2), columns=['a', 'b'])
noise = np.random.randn(100)
df['c'] = (3*df['a']) + (2*df['b']) + (.5*noise)
vifs = variance_inflation_factors(df)
assert type(vifs) == pd.Series
assert vifs.max() > 5.0
assert vifs.idxmax() == 'c'
def test_flagging():
df = pd.DataFrame(np.random.randn(100, 2), columns=['a', 'b'])
noise = np.random.randn(100)
df['c'] = (3*df['a']) + (2*df['b']) + (.5*noise)
diagnostics = Diagnostics(df)
rows, cols = diagnostics.flag_all({'VIFs' : (lambda x: x > 0),
'RowMahalanobisDistances' : (lambda x: x > 0)})
# Everything should be flagged
assert np.array_equal(rows, range(df.shape[0]))
assert np.array_equal(cols, range(df.shape[1]))
vif = variance_inflation_factors(df).max()
rows, cols = diagnostics.flag_all({'VIFs' : (lambda x: x >= vif)},
include=['VIFs'])
assert np.array_equal(rows, [])
assert np.array_equal(cols, [2])
rows, cols = diagnostics.flag_all(exclude=['VIFs', 'ConditionIndices',
'Eigenvalues', 'CorrelationMatrix',
'RowMahalanobisDistances', 'ColumnMahalanobisDistances',
'Variances'])
assert np.array_equal(rows, [])
assert np.array_equal(cols, [])