"""
created matt_dumont
on: 3/10/23
"""
import pandas as pd
from komanawa.gw_detect_power.change_detection_slope import DetectionPowerSlope
from komanawa.kendall_stats import example_data
import numpy as np
import itertools
from pathlib import Path
import timeit
import sys
import os
# constants
nsims = 10
mpmk_check_step = 1
mpmk_efficent_min = 10
mpmk_window = 0.05
nsims_pettit = 2000
# iterables
methods = DetectionPowerSlope.implemented_significance_modes
ndata = (50, 100, 500, 1000, 5000)
efficency_modes = (True, False)
[docs]
def timeit_test(methods=methods, ndata=ndata, efficency_modes=efficency_modes, n=1):
"""
run an automated timeit test, must be outside of the function definition, prints results in scientific notation
units are seconds
:param methods: list of methods to test
:param ndata: list of data sizes to test
:param efficency_modes: list of efficency modes to test
:param n: number of times to test
:return:
"""
py_file_path = __file__
print(py_file_path)
d = os.path.dirname(py_file_path)
fname = os.path.basename(py_file_path).replace('.py', '')
sys.path.append(d)
out = pd.DataFrame(index=pd.MultiIndex.from_product([ndata, efficency_modes],
names=['n data', 'efficency_mode']),
columns=methods)
for nd, emode, method in itertools.product(ndata, efficency_modes, methods):
print(f'testing: {method=}, {nd=}, {emode=}')
fn = f'run_model'
t = timeit.timeit(f'{fn}("{method}", {nd}, {emode})',
setup='from {} import {}'.format(fname, fn),
number=n) / n
out.loc[(nd, emode), method] = t
print('{0:e} seconds'.format(t))
return out
def run_model(method, ndata, emode):
if method in ['linear-regression', 'mann-kendall', ]:
x, data = example_data.make_increasing_decreasing_data(slope=0.1, noise=0, step=100 / ndata)
use_noise = 5
elif method in ['linear-regression-from-max', 'mann-kendall-from-max', 'n-section-mann-kendall']:
x, data = example_data.make_multipart_sharp_change_data(slope=example_data.multipart_sharp_slopes[0],
noise=0,
unsort=False, na_data=False, step=100 / ndata)
use_noise = example_data.multipart_sharp_noises[1]
elif method in ['linear-regression-from-min', 'mann-kendall-from-min', ]:
x, data = example_data.make_multipart_sharp_change_data(slope=example_data.multipart_sharp_slopes[1],
noise=0,
unsort=False, na_data=False, step=100 / ndata)
use_noise = example_data.multipart_sharp_noises[1]
elif method == 'pettitt-test':
data = np.zeros((ndata)) + 10
data[len(data) * 3 // 4:] = 8
use_noise = 5
else:
raise ValueError(f'unknown method {method}')
expect_slope = 'auto'
if method == 'n-section-mann-kendall':
expect_slope = (1, -1)
dpc = DetectionPowerSlope(significance_mode=method,
nsims=nsims, min_p_value=0.05, min_samples=10,
expect_slope=expect_slope, efficent_mode=emode, nparts=2, min_part_size=10,
no_trend_alpha=0.50,
mpmk_check_step=mpmk_check_step, mpmk_efficent_min=mpmk_efficent_min,
mpmk_window=mpmk_window,
nsims_pettit=nsims_pettit, )
out = dpc.power_calc(idv='test',
error=use_noise,
mrt_model='pass_true_conc',
true_conc_ts=data)
if __name__ == '__main__':
data = timeit_test()
data.to_csv(Path(__file__).parent.joinpath('timeit_test_results.txt'))
