def find_reference_dispersion(data, k, number_of_bootstraps=10):
"""
Finds the reference dispersion (and confidence) for the data supplied.
"""
incremental_statistic = IncrementalStatistic()
logging.info(
'Finding {} reference dispersions'.format(number_of_bootstraps))
for run_number in range(number_of_bootstraps):
start = datetime.datetime.utcnow()
logging.info('At iteration {}'.format(run_number))
# uniform_points =
# generate_principal_components_box_uniform_points(data)
uniform_points = generate_bounding_box_uniform_points(data)
dispersion, _, _ = default_clustering(uniform_points, k, 1, 500)
if 0 == dispersion:
logging.warning(
'[Reference Dispersion] Cannot take the log of 0 for run '
'number = {}.'.format(run_number))
continue
incremental_statistic.add_value(np.log(dispersion))
end = datetime.datetime.utcnow()
logging.info('Time for last reference set: {}'.format((end - start)))
stddev_dispersions = incremental_statistic.get_standard_deviation() / \
np.sqrt(1 + 1 / number_of_bootstraps)
return incremental_statistic.get_mean(), stddev_dispersions
