def test_scale(collection):
transform.Scale(collection, variables=['RT', 'parametric gain'],
output=['RT_Z', 'gain_Z'], groupby=['run', 'subject'])
groupby = collection['RT'].get_grouper(['run', 'subject'])
z1 = collection['RT_Z'].values
z2 = collection['RT'].values.groupby(
groupby).apply(lambda x: (x - x.mean()) / x.std())
assert np.allclose(z1, z2)
def test_scale(collection, sparse_run_variable_with_missing_values):
transform.Scale(collection, variables=['RT', 'parametric gain'],
output=['RT_Z', 'gain_Z'], groupby=['run', 'subject'])
groupby = collection['RT'].get_grouper(['run', 'subject'])
z1 = collection['RT_Z'].values
z2 = collection['RT'].values.groupby(
groupby).apply(lambda x: (x - x.mean()) / x.std())
assert np.allclose(z1, z2)
# Test constant input
coll = sparse_run_variable_with_missing_values
coll['var'].values.fillna(1)
with pytest.raises(ValueError, match="Cannot scale.*1.0"):
transform.Scale(coll, 'var').values
with pytest.raises(ValueError, match="Cannot scale.*1.0"):
transform.Scale(coll, 'var', replace_na='before').values
transform.Scale(coll, 'var', replace_na='after', output='zero')
assert coll['zero'].values.unique() == 0