[Python] Running a non-parametric stats analysis using a python jupyterhub script, but i have...

I am looking into comparing VCL as an outcome response in 3 different incubation times (1h, 2h, 3h) and 4 treatment conditions (LT, LC, HTF, MRS).

The boxplots show a clear difference between LC and other treatments at all time points, but I am having issues running the stats script to compare LC and other treatment conditions at specific timings. See code below

import pandas as pd
from scipy.stats import mannwhitneyu
from itertools import combinations
from statsmodels.stats.multitest import multipletests

# Clean column names if needed
df1.columns = df1.columns.str.strip().str.replace(' ', '_')

# Clean string values
df1['Treatment'] = df1['Treatment'].str.strip()
df1['Incubation_time'] = df1['Incubation_time'].astype(str).str.strip()

# Step 1: Exclude 0H incubation
df_no0H = df1[df1['Incubation_time'] != '0H']

# Step 2: Subset to only 1H time
df_1H = df_no0H[df_no0H['Incubation_time'] == '1H']

# Step 3: Prepare VCL data for 4 treatment groups
treatments = ['LT', 'LC', 'HTF', 'MRS']
group_data = {
t: df_1H[df_1H['Treatment'] == t]['VCL'].dropna().values
for t in treatments
}

# Step 4: Mann-Whitney U test for all pairs
results = []
pairs = list(combinations(treatments, 2))

for a, b in pairs:
if len(group_data[a]) >= 3 and len(group_data) >= 3:
stat, p = mannwhitneyu(group_data[a], group_data, alternative='two-sided')
results.append({'Comparison': f'{a} vs {b}', 'U': stat, 'p-value': p})
else:
results.append({'Comparison': f'{a} vs {b}', 'U': None, 'p-value': None})

# Step 5: Bonferroni correction
pvals = [r['p-value'] for r in results if r['p-value'] is not None]
_, corrected, _, _ = multipletests(pvals, method='bonferroni')

# Add corrected p-values back into results
i = 0
for r in results:
if r['p-value'] is not None:
r['p-value (Bonferroni)'] = corrected
i += 1
else:
r['p-value (Bonferroni)'] = None

# Step 6: Convert to DataFrame for display
results_df = pd.DataFrame(results)
print(results_df)

Continue reading...