Calculate statistical significance

twitter_saliency_filter_analysis--14.py

w, p = wilcoxon(pairwise_df['baseline_h2']-pairwise_df['saliency_out'])
print(w,p)
pairwise_df['globalh2_wt_p'] = p
pairwise_df['globalh2_wt_w'] = w
w, p = wilcoxon(pairwise_df['baseline_h1']-pairwise_df['saliency_out'])
print(w,p)
pairwise_df['globalh1_wt_p'] = p
pairwise_df['globalh1_wt_w'] = w
w, p = wilcoxon(pairwise_df['baseline_v2']-pairwise_df['saliency_out'])
print(w,p)
pairwise_df['globalv2_wt_p'] = p
pairwise_df['globalv2_wt_w'] = w
w, p = wilcoxon(list(pairwise_df['baseline_v1']-pairwise_df['saliency_out']))
print(w,p)
pairwise_df['globalv1_wt_p'] = p
pairwise_df['globalv1_wt_w'] = w

pairwise_df['localh2_wt_p'] = np.nan
pairwise_df['localh2_wt_w'] = np.nan
pairwise_df['localh1_wt_p'] = np.nan
pairwise_df['localh1_wt_w'] = np.nan
pairwise_df['localv2_wt_p'] = np.nan
pairwise_df['localv2_wt_w'] = np.nan
pairwise_df['localv1_wt_p'] = np.nan
pairwise_df['localv1_wt_w'] = np.nan

for expID in tqdm(list(set(list(pairwise_df.experiment_id.values)))):
  condition = pairwise_df['experiment_id'] == expID

  diff = list(pairwise_df.loc[condition,['baseline_h2']].values-pairwise_df.loc[condition,['saliency_out']].values)
  diff = [list(d)[0]for d in diff]
  try:
    w, p = wilcoxon(diff)
    pairwise_df.loc[condition,'localh2_wt_p'] = p
    pairwise_df.loc[condition,'localh2_wt_w'] = w
  except ValueError as e:
    print(f'Skipping Wilcoxon Signed Rank test for: {expID} due to: \n{e}')

  diff = list(pairwise_df.loc[condition,['baseline_h1']].values-pairwise_df.loc[condition,['saliency_out']].values)
  diff = [list(d)[0]for d in diff]
  try:
    w, p = wilcoxon(diff)
    pairwise_df.loc[condition,'localh1_wt_p'] = p
    pairwise_df.loc[condition,'localh1_wt_w'] = w
  except ValueError as e:
    print(f'Skipping Wilcoxon Signed Rank test for: {expID} due to: \n{e}')

  diff = list(pairwise_df.loc[condition,['baseline_v2']].values-pairwise_df.loc[condition,['saliency_out']].values)
  diff = [list(d)[0]for d in diff]
  try:
    w, p = wilcoxon(diff)
    pairwise_df.loc[condition,'localv2_wt_p'] = p
    pairwise_df.loc[condition,'localv2_wt_w'] = w
  except ValueError as e:
    print(f'Skipping Wilcoxon Signed Rank test for: {expID} due to: \n{e}')

  diff = list(pairwise_df.loc[condition,['baseline_v1']].values-pairwise_df.loc[condition,['saliency_out']].values)
  diff = [list(d)[0]for d in diff]
  try:
    w, p = wilcoxon(diff)
    pairwise_df.loc[condition,'localv1_wt_p'] = p
    pairwise_df.loc[condition,'localv1_wt_w'] = w
  except ValueError as e:
    print(f'Skipping Wilcoxon Signed Rank test for: {expID} due to: \n{e}')