def collection():
"""Create a collection at dataset level."""
df = pd.DataFrame({
'subject': np.repeat(np.arange(10), 4),
'run': np.tile(np.arange(4), 10),
'age': np.repeat(np.random.normal(40, 10, 10).astype(int), 4)
})
entities = df[['subject', 'run']]
return BIDSVariableCollection.from_df(df, entities=entities)
def _concatenate_input_nodes(self, nodes):
data, entities = [], []
for n in nodes:
contrasts = [c.name for c in n.contrasts]
row = pd.Series(np.ones(len(contrasts)), index=contrasts)
data.append(row)
entities.append(pd.Series(n.entities))
data = pd.concat(data, axis=1, sort=True).T
entities = pd.concat(entities, axis=1, sort=True).T
return BIDSVariableCollection.from_df(data, entities, self.level)
def _concatenate_input_nodes(self, nodes):
data, entities = [], []
for n in nodes:
contrasts = [c.name for c in n.contrasts]
row = pd.Series(np.ones(len(contrasts)), index=contrasts)
data.append(row)
entities.append(pd.Series(n.entities))
data = pd.concat(data, axis=1, sort=True).T
entities = pd.concat(entities, axis=1, sort=True).T
return BIDSVariableCollection.from_df(data, entities, self.level)
def _merge_contrast_inputs(self, inputs):
""" Merges a list of ContrastInfo tuples and constructs a dict mapping
from units of the current level to BIDSVariableCollections.
Parameters
----------
inputs: [[ContrastInfo]]
List of list of ContrastInfo tuples passed from the previous Step.
Each element in the outer list maps to the output of a unit at the
previous level; each element in the inner list is a ContrastInfo
tuple. E.g., if contrast information is being passed from run-level
to subject-level, each outer element is a run.
Returns
-------
A dictionary, where the keys are the values of the entities at the
current level (e.g., '01', '02'...) and the values are
BIDSVariableCollection containing contrast information.
Notes
-----
Each output BIDSVariableCollection contains information for a single
unit at the present level. The variables in the collection reflect the
union of all contrasts found in one or more of the inputs. A value of
1 indicates that the contrast is present for a given row in the input;
0 indicates that the contrast was missing.
"""
groups = self._group_objects_by_entities(inputs)
ent_cols = list(list(groups.values())[0][0].entities.keys())
collections = {}
for name, contrasts in groups.items():
# Create a DF with contrasts in rows and contrast names in columns
data = [{**c.entities, **{c.name: 1}} for c in contrasts]
data = pd.DataFrame.from_records(data)
# Group by all entities and sum, collapsing over rows belonging
# to the current unit
data = data.groupby(ent_cols).sum()
# Split the DF up into separate data and entities DFs
entities = data.index.to_frame(index=False)
data = data.reset_index(drop=True)
# Construct the collection
coll = BIDSVariableCollection.from_df(data, entities, self.level)
collections[name] = coll
return collections