The Toolkit

Ten p-values, eight of them below 0.05. Bonferroni keeps 2 as significant; the Benjamini–Hochberg procedure keeps 6, controlling the expected share of false discoveries instead of the chance of any. statsmodels’ multipletests does both; which one is right depends on whether a single false positive is costly or a few are acceptable.

• Corrections apply to the family of tests you actually ran, including the ones you did not report.
• Benjamini–Hochberg assumes independent or positively dependent tests; under other dependence use Benjamini–Yekutieli.
• A non-significant result after correction is not evidence of no effect.

p = np.array([0.001, 0.004, 0.008, 0.012, 0.02, 0.03, 0.04, 0.045, 0.2, 0.5])
m = len(p)
bonferroni = (p < 0.05 / m).sum()
order = np.sort(p)
passed = order <= 0.05 * np.arange(1, m + 1) / m    # Benjamini-Hochberg step-up
bh = (np.max(np.nonzero(passed)) + 1) if passed.any() else 0

Assumes import numpy as np and import pandas as pd (and SciPy where imported). Each snippet simulates or makes up its own data, as the comments say.

# scipy.stats — statistical tests
from scipy.stats import (
    ttest_ind, ttest_rel, mannwhitneyu,
    ks_2samp, shapiro, spearmanr
)

# Two-sample t-test
t_stat, p_val = ttest_ind(group_a, group_b)

# Normality check
stat, p = shapiro(data)

# statsmodels — regression with full diagnostics
import statsmodels.api as sm

X_sm = sm.add_constant(X)
model = sm.OLS(y, X_sm).fit()
print(model.summary())  # R², coefficients, p-values, CI

# Time series — ADF stationarity test
from statsmodels.tsa.stattools import adfuller
result = adfuller(series)
print(f"ADF Stat: {result[0]:.3f}, p: {result[1]:.4f}")