wilcoxon.py

# TODO: refactor for future usage
# TODO: Bonferroni-correct for number of questions

import pandas as pd
import scikit_posthocs as sp
import statsmodels.api as sa

def load_questionnaire():
    q = pd.read_csv("questionnaire.csv", delimiter=";")
    q = q.melt(id_vars=["Technique", "QuestionId"], value_vars=[f"P{i}" for i in range(1, 13)], var_name="ParticipantId", value_name="Response")
    return q

def test_question(question_id, questionnaire):
        q_samples = questionnaire[questionnaire["QuestionId"] == question_id] 
        test_result = sp.posthoc_wilcoxon(q_samples, val_col="Response", group_col="Technique", p_adjust="bonferroni")
        return pd.DataFrame(test_result)

def test_all_questions(alpha):
    questionnaire = load_questionnaire()
    interesting_questions = {}
    for q_id in questionnaire["QuestionId"].unique():
        test_result = test_question(q_id, questionnaire)
        if True in (test_result < alpha).values:
            interesting_questions[q_id] = test_result
    return interesting_questions
  
q = load_questionnaire()
qs = test_all_questions(alpha=.05)
qs