path='average' concatenates the result using wrong index

[hbaniecki]

mean isn't calculated over variable but path orders (pd.concat uses integer index to join, not variable)

DALEX/python/dalex/dalex/instance_level/_shap/utils.py

Lines 36 to 50 in b0ca23c

	if path is not None:
	if isinstance(path, str) and path == 'average':
	# let's calculate an average attribution
	extracted_contributions = [None] * B
	for i in range(B):
	extracted_contributions[i] = result[i].sort_values(['label', 'variable']).loc[:, 'contribution']
	extracted_contributions = pd.concat(extracted_contributions, axis=1)
	result_average = deepcopy(result[0])
	result_average = result_average.sort_values(['label', 'variable'])
	result_average['contribution'] = extracted_contributions.mean(axis=1)
	result_average['B'] = 0
	result_average['sign'] = np.sign(result_average['contribution'])
	result.append(result_average)