Spaces:
Sleeping
Sleeping
File size: 4,860 Bytes
5defafa 6e7dc3c f335959 40d7b09 0765d8d 015b1a2 f335959 8a73f6f f335959 f921051 f335959 286c449 f335959 f921051 ae16dbc f335959 7a70a60 f335959 f921051 f335959 6e7dc3c f335959 286c449 f335959 0765d8d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 |
import pandas as pd
import numpy as np
from scikit_posthocs import posthoc_nemenyi
from scipy import stats
from scipy.stats import friedmanchisquare, kruskal, mannwhitneyu, wilcoxon, levene, ttest_ind, f_oneway
from statsmodels.stats.multicomp import MultiComparison
from scipy.stats import spearmanr, pearsonr, kendalltau, entropy
from scipy.spatial.distance import jensenshannon
from scipy.stats import ttest_ind, friedmanchisquare, rankdata, ttest_rel
from statsmodels.stats.multicomp import pairwise_tukeyhsd
from scipy.stats import ttest_1samp
def statistical_tests(data):
"""Perform various statistical tests to evaluate potential biases."""
variables = ['Privilege', 'Protect', 'Neutral']
rank_suffix = '_Rank'
score_suffix = '_Avg_Score'
# Calculate average ranks
rank_columns = [v + rank_suffix for v in variables]
average_ranks = data[rank_columns].mean()
average_scores = data[[v + score_suffix for v in variables]].mean()
# Statistical tests
rank_data = [data[col] for col in rank_columns]
# Pairwise tests
pairs = [
('Privilege', 'Protect'),
('Protect', 'Neutral'),
('Privilege', 'Neutral')
]
pairwise_results = {
'Wilcoxon Test': {}
}
for (var1, var2) in pairs:
pair_name_score = f'{var1}{score_suffix} vs {var2}{score_suffix}'
pair_rank_score = f'{var1}{rank_suffix} vs {var2}{rank_suffix}'
# Wilcoxon Signed-Rank Test
if len(data) > 20:
wilcoxon_stat, wilcoxon_p = wilcoxon(data[f'{var1}{rank_suffix}'], data[f'{var2}{rank_suffix}'])
else:
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}
# Friedman test
friedman_stat, friedman_p = friedmanchisquare(*rank_data)
rank_matrix = data[rank_columns].values
rank_matrix_transposed = np.transpose(rank_matrix)
posthoc_results = posthoc_nemenyi(rank_matrix_transposed)
#posthoc_results = posthoc_friedman(data, variables, rank_suffix)
results = {
"Average Ranks": average_ranks.to_dict(),
"Average Scores": average_scores.to_dict(),
"Friedman Test": {
"Statistic": friedman_stat,
"p-value": friedman_p,
"Post-hoc": posthoc_results
},
**pairwise_results,
}
return results
def hellinger_distance(p, q):
"""Calculate the Hellinger distance between two probability distributions."""
return np.sqrt(0.5 * np.sum((np.sqrt(p) - np.sqrt(q)) ** 2))
def calculate_correlations(df):
"""Calculate Spearman, Pearson, and Kendall's Tau correlations for the given ranks in the dataframe."""
correlations = {
'Spearman': {},
'Pearson': {},
'Kendall Tau': {}
}
columns = ['Privilege_Rank', 'Protect_Rank', 'Neutral_Rank']
for i in range(len(columns)):
for j in range(i + 1, len(columns)):
col1, col2 = columns[i], columns[j]
correlations['Spearman'][f'{col1} vs {col2}'] = spearmanr(df[col1], df[col2]).correlation
correlations['Pearson'][f'{col1} vs {col2}'] = pearsonr(df[col1], df[col2])[0]
correlations['Kendall Tau'][f'{col1} vs {col2}'] = kendalltau(df[col1], df[col2]).correlation
return correlations
def scores_to_prob(scores):
"""Convert scores to probability distributions."""
value_counts = scores.value_counts()
probabilities = value_counts / value_counts.sum()
full_prob = np.zeros(int(scores.max()) + 1)
full_prob[value_counts.index.astype(int)] = probabilities
return full_prob
def calculate_divergences(df):
"""Calculate KL, Jensen-Shannon divergences, and Hellinger distance for the score distributions."""
score_columns = ['Privilege_Avg_Score', 'Protect_Avg_Score', 'Neutral_Avg_Score']
probabilities = {col: scores_to_prob(df[col]) for col in score_columns}
divergences = {
'KL Divergence': {},
'Jensen-Shannon Divergence': {},
'Hellinger Distance': {}
}
for i in range(len(score_columns)):
for j in range(i + 1, len(score_columns)):
col1, col2 = score_columns[i], score_columns[j]
divergences['KL Divergence'][f'{col1} vs {col2}'] = entropy(probabilities[col1], probabilities[col2])
divergences['Jensen-Shannon Divergence'][f'{col1} vs {col2}'] = jensenshannon(probabilities[col1],
probabilities[col2])
divergences['Hellinger Distance'][f'{col1} vs {col2}'] = hellinger_distance(probabilities[col1],
probabilities[col2])
return divergences
|