def test_resampler_noreplace(df):
np.random.seed(28)
resampler = iters.Resample(df, 'ID', replace=False, sample_size=3)
next(resampler)
resampler = iters.Resample(df, 'ID', stratify='STRAT')
(new_df, ids) = next(resampler)
assert ids == [1, 2, 4]
assert list(new_df['ID']) == [1, 1, 2, 2, 3, 3]
assert list(new_df['DV']) == [5, 6, 3, 4, 0, 9]
resampler = iters.Resample(df, 'ID', replace=False, sample_size=2)
(new_df, ids) = next(resampler)
assert list(ids) == [2, 1]
assert list(new_df['ID']) == [1, 1, 2, 2]
def test_resampler_default(df):
np.random.seed(28)
resampler = iters.Resample(df, 'ID')
(new_df, ids) = next(resampler)
assert ids == [1, 4, 2]
assert list(new_df['ID']) == [1, 1, 2, 2, 3, 3]
assert list(new_df['DV']) == [5, 6, 0, 9, 3, 4]
assert list(new_df['STRAT']) == [1, 1, 2, 2, 2, 2]
assert new_df.name == 'resample_1'
with pytest.raises(StopIteration): # Test the default one iteration
next(resampler)
def test_resampler_anonymization(testdata):
np.random.seed(28)
df = pharmpy.data.read_csv(testdata / 'pheno_data.csv')
resampler = iters.Resample(df, group='ID')
(new_df, ids) = next(resampler)
assert all(e in ids for e in range(1, 60))
assert len(ids) == 59
for new_id, old_id in enumerate(ids, start=1):
df_newid = new_df.query(f'ID == {new_id}')
df_oldid = df.query(f'ID == {old_id}')
df_newid = df_newid.reset_index(drop=True)
df_oldid.reset_index(inplace=True, drop=True)
df_newid['ID'] = old_id
pandas.testing.assert_frame_equal(df_newid, df_oldid)
def test_stratification(df):
np.random.seed(28)
resampler = iters.Resample(df,
'ID',
resamples=1,
stratify='STRAT',
sample_size={
1: 2,
2: 3
},
replace=True)
(new_df, ids) = next(resampler)
assert ids == [1, 1, 4, 4, 4]
assert list(new_df['ID']) == [1, 1, 2, 2, 3, 3, 4, 4, 5, 5]
assert list(new_df['DV']) == [5, 6, 5, 6, 0, 9, 0, 9, 0, 9]
resampler = iters.Resample(df,
'ID',
resamples=1,
stratify='STRAT',
sample_size={1: 2},
replace=True)
(new_df, ids) = next(resampler)
assert ids == [1, 1]
resampler = iters.Resample(df,
'ID',
resamples=3,
stratify='STRAT',
replace=True)
(new_df, ids) = next(resampler)
assert ids == [1, 2, 2]
(new_df, ids) = next(resampler)
assert ids == [1, 2, 4]
(new_df, ids) = next(resampler)
assert ids == [1, 4, 2]
def test_resampler_too_big_sample_size(df):
with pytest.raises(ValueError):
iters.Resample(df, 'ID', replace=False, sample_size=4)