hypothesis-testing-tool 0.1.2

A tool for doing hypothesis testing

import sys
import os
sys.path.append(os.path.abspath(os.path.join('..')))

Description

A package to run hypothesis testing for one and two samples.

One Sample Hypothesis Testing

from hypothesis_testing_tool.compute_stats.one_sample_statistics import OneSampleTest

Let's see an example of how you can use the OneSampleTest class.

import random

# Set the seed for reproducibility
seed_value = 42
random.seed(seed_value)

random_sample = [random.gauss(mu = 5, sigma = 1) for _ in range(1000)]

t_test = OneSampleTest(
        data = random_sample,
        null_population_mean = 3.5
        ).t_test_results

print(f"p-value: {t_test.pvalue:.2f}")
print(f"t-statistic: {t_test.statistic:.2f}")

p-value: 0.00
t-statistic: 45.96

The default is with alternative = "two-sided", but that can change to a one tail test.

t_test_less = OneSampleTest(
        data = random_sample,
        null_population_mean = 3.5,
        alternative = "less"
        ).t_test_results

print(f"p-value: {t_test_less.pvalue:.2f}")
print(f"t-statistic: {t_test_less.statistic:.2f}")

p-value: 1.00
t-statistic: 45.96

t_test_greater = OneSampleTest(
        data = random_sample,
        null_population_mean = 3.5,
        alternative = "greater"
        ).t_test_results

print(f"p-value: {t_test_greater.pvalue:.2f}")
print(f"t-statistic: {t_test_greater.statistic:.2f}")

p-value: 0.00
t-statistic: 45.96

You can also compute the confidence interval (default = 95%) for the mean, using the calculate_ci method.

The calculate_ci method takes an optional argument alpha to adjust to 99% CI (alpha = 0.01) or any other.

confidence_interval = OneSampleTest(
        data = random_sample,
        null_population_mean = 3.5
        ).calculate_ci()
confidence_interval

{'lower_bound': 4.918459428969473,
 'point_estimate': 4.9817306915192265,
 'upper_bound': 5.04500195406898,
 'null_population_mean': 3.5}

Finally, you can create a plot with the CI and save it to a local path.

from hypothesis_testing_tool.presentation.create_plots import create_one_sample_hypothesis_plot

create_one_sample_hypothesis_plot(
    path_to_save_plot = "artifacts/one_sample_plot.png",
    ci_dict = confidence_interval,
    width = 10,
    height = 5
)

(<Figure size 1000x500 with 1 Axes>,
 <Axes: title={'center': '95% Confidence Interval (CI) for the Mean of One Sample'}, xlabel='Values'>)

In the plot above the 95% confidence interval includes 3.5, so we do not reject the null hypothesis.