def _get_multi_choice_data(
self, question_metadata: QuestionMetadata
) -> Series:
# merge multi-choice questions to single column
level_0 = self.survey_data.columns.get_level_values(0).to_list()
level_1 = self.survey_data.columns.get_level_values(1).to_list()
if question_metadata.expression is None:
match_cols = [(l0, l1) for l0, l1 in zip(level_0, level_1)
if l0 == question_metadata.text]
else:
match_cols = [(l0, l1) for l0, l1 in zip(level_0, level_1)
if match(question_metadata.expression, l0)]
if len(match_cols) == 0:
if question_metadata.expression is None:
raise ValueError(
f'Could not find any columns matching '
f'"{question_metadata.text}"'
)
else:
raise ValueError(
f'Could not match expression '
f'"{question_metadata.expression}" '
f'for MultiChoice question "{question_metadata}"'
)
return Series(
name=match_cols[0][0],
index=self.survey_data.index,
data=self.survey_data[match_cols].apply(
lambda row: CATEGORY_SPLITTER.join(
row.dropna().astype(str)
) if len(row.dropna()) else nan,
axis=1)
)
def replace_item(item: str):
if isnull(item):
return nan
items = item.split(CATEGORY_SPLITTER)
items = [
replacements[item] if item in replacements.keys() else item
for item in items
]
return CATEGORY_SPLITTER.join(items)
def rank_choices(value: str):
choices_ranks = value.split(' | ')
choices = [rank_choice.split(':')[0]
for rank_choice in choices_ranks]
ranks = [int(rank_choice.split(':')[1])
for rank_choice in choices_ranks]
ranked_choices = [choice for rank, choice in
sorted(zip(ranks, choices))]
return CATEGORY_SPLITTER.join(
str(choice) for choice in ranked_choices
)
def get_selected(item: str):
if isnull(item):
return nan
selected = []
while len(item) > 0:
for category in categories:
if item == category:
selected.append(category)
item = ''
elif item[: len(category) + 2] == category + '; ':
selected.append(category)
item = item[len(category) + 2:]
return CATEGORY_SPLITTER.join(selected)
def _get_multi_choice_data(self,
question_metadata: QuestionMetadata) -> Series:
# merge multi-choice questions to single column
if question_metadata.expression is None:
raise ValueError('Need a regular expression to match '
'MultiChoice question columns.')
match_cols = [
c for c in self.survey_data.columns
if match(question_metadata.expression, c)
]
if len(match_cols) == 0:
raise ValueError(
f'Could not match expression "{question_metadata.expression}" '
f'for MultiChoice question "{question_metadata}"')
return self.survey_data[match_cols].apply(
lambda row: CATEGORY_SPLITTER.join(row.dropna().astype(str)),
axis=1)
def _get_multi_choice_data(
self, question_metadata: QuestionMetadata
) -> Series:
# get data
data: DataFrame = self.survey_data[question_metadata.text]
# replace other values
if question_metadata.text in self.questions_metadata['text'].to_list():
# this if condition excludes repeated questions that have had their
# text changed
def replace_other(value):
if isnull(value):
return value
else:
if value in categories:
return value
else:
return 'Other'
metadata: Series = self.questions_metadata.loc[
self.questions_metadata['text'] == question_metadata.text
].iloc[0]
if 'other' in metadata.keys() and metadata['other'] == True:
# do replacement
category_name = metadata['categories']
categories = self.orders_metadata.loc[
self.orders_metadata['category'] == category_name,
'value'
].to_list()
data = data.applymap(replace_other)
# update categories in orders_metadata
self.orders_metadata = self.orders_metadata.append(
Series({'category': question_metadata.name,
'value': 'Other'}),
ignore_index=True
)
# merge multi-choice questions to single column
return data.apply(
lambda row: CATEGORY_SPLITTER.join(row.dropna().astype(str)),
axis=1
)
def _get_ranked_choice_data(self,
question_metadata: QuestionMetadata) -> list:
# merge ranked choice questions to single column
if question_metadata.expression is None:
raise ValueError('Need a regular expression to match '
'RankedChoice question columns.')
match_cols = [
c for c in self.survey_data.columns
if match(question_metadata.expression, c)
]
choices = [self._get_ranked_choice_name(c) for c in match_cols]
# create column in new dataframe
new_answers = []
for _, row in self.survey_data[match_cols].iterrows():
# order choices by rank
ranks = row.tolist()
ranked_choices = [
choice for rank, choice in sorted(zip(ranks, choices))
]
ranked_choices = CATEGORY_SPLITTER.join(
str(choice) for choice in ranked_choices)
new_answers.append(ranked_choices)
return new_answers
def stack(self, name: str,
drop_na: bool = True,
null_category: Optional[str] = None,
name_index: Optional[str] = None,
key_index: Optional[str] = None,
number_index: Optional[str] = None,
**kwargs):
"""
Stack the responses to each question in the group into a new question.
:param name: Name for the new question.
:param drop_na: For MultiChoiceQuestions, whether to drop rows where the
respondent was not asked the question.
:param null_category: For MultiChoiceQuestions, optional response to
exclude from the new question group.
:param name_index: Name of a new index column to create with values
corresponding to the name of the question the data
comes from.
:param key_index: Name of a new index column to create with values
corresponding to the question's key in the group.
:param number_index: Name of a new index column to create with values
corresponding to the number of the question the
data comes from.
:param kwargs: Optional new attribute values to override in the new
question.
"""
if name == '':
raise ValueError('Name cannot be empty.')
# create index names
index_names = [self._questions[0].data.index.name]
if name_index is not None:
index_names.append(name_index)
if key_index is not None:
index_names.append(key_index)
if number_index is not None:
index_names.append(number_index)
question_datas = []
question: MultiChoiceQuestion
for question in self._questions:
# create data
question_data = question.make_features(
naming='{{choice}}', drop_na=drop_na,
null_category=null_category
)
# create index
index_list = question_data.index.to_list()
if name_index is not None:
name_list = [question.name] * len(question_data)
else:
name_list = None
if key_index is not None:
key_list = [
[k for k in self._item_dict.keys()
if self._item_dict[k] == question][0]
] * len(question_data)
else:
key_list = None
if number_index is not None:
number_list = [
self._questions.index(question)
] * len(question_data)
else:
number_list = None
if name_list is None and key_list is None and number_list is None:
question_data.index = Index(data=index_list,
name=index_names[0])
else:
index_tuples = list(zip(*[
ix_list for ix_list in [index_list, name_list,
key_list, number_list]
if ix_list is not None
]))
question_data.index = MultiIndex.from_tuples(
tuples=index_tuples, names=index_names
)
question_datas.append(question_data)
new_data = concat(question_datas, axis=0)
new_data = DataFrame({
column: column_data.replace({1: column, 0: ''})
for column, column_data in new_data.iteritems()
})
new_data = Series(
data=[
CATEGORY_SPLITTER.join(row.replace('', nan).dropna().to_list())
for _, row in new_data.iterrows()
],
index=new_data.index, name=name
)
# copy question
new_question = copy(self._questions[0])
new_question.name = name
new_question._data = new_data
if isinstance(new_question, MultiChoiceQuestion):
if null_category is not None:
new_question._categories = [
c for c in new_question._categories
if c != null_category
]
for kw, arg in kwargs.items():
setattr(new_question, kw, arg)
return new_question