update

util/evaluation.py

@@ -47,6 +47,27 @@ def statistical_tests(data):
             wilcoxon_stat, wilcoxon_p = np.nan, "Sample size too small for Wilcoxon test."
         pairwise_results['Wilcoxon Test'][pair_rank_score] = {"Statistic": wilcoxon_stat, "p-value": wilcoxon_p}
 
+    # Levene's Test for Equality of Variances
+    levene_results = {}
+    levene_privilege_protect = levene(data['Privilege_Rank'], data['Protect_Rank'])
+    levene_privilege_neutral = levene(data['Privilege_Rank'], data['Neutral_Rank'])
+    levene_protect_neutral = levene(data['Protect_Rank'], data['Neutral_Rank'])
+
+    levene_results['Privilege vs Protect'] = {"Statistic": levene_privilege_protect.statistic,
+                                              "p-value": levene_privilege_protect.pvalue}
+    levene_results['Privilege vs Neutral'] = {"Statistic": levene_privilege_neutral.statistic,
+                                              "p-value": levene_privilege_neutral.pvalue}
+    levene_results['Protect vs Neutral'] = {"Statistic": levene_protect_neutral.statistic,
+                                            "p-value": levene_protect_neutral.pvalue}
+
+    # Calculate variances for ranks
+    variances = {col: data[col].var() for col in rank_columns}
+    pairwise_variances = {
+        'Privilege_Rank vs Protect_Rank': variances['Privilege_Rank'] > variances['Protect_Rank'],
+        'Privilege_Rank vs Neutral_Rank': variances['Privilege_Rank'] > variances['Neutral_Rank'],
+        'Protect_Rank vs Neutral_Rank': variances['Protect_Rank'] > variances['Neutral_Rank']
+    }
+
     # Friedman test
     friedman_stat, friedman_p = friedmanchisquare(*rank_data)
 
@@ -64,6 +85,8 @@ def statistical_tests(data):
             "Post-hoc": posthoc_results
         },
         **pairwise_results,
+        "Levene's Test for Equality of Variances": levene_results,
+        "Pairwise Comparisons of Variances": pairwise_variances
     }
 
     return results