PageRenderTime 56ms CodeModel.GetById 17ms RepoModel.GetById 0ms app.codeStats 0ms

/AUC/AUC_plot.py

https://gitlab.com/BudzikFUW/budzik_analiza
Python | 228 lines | 148 code | 38 blank | 42 comment | 15 complexity | 612fcd80870bd3e3c1d188f798170b24 MD5 | raw file
    

  1. 

  2. import pandas as pd
    3. 

  3. import numpy as np
    4. 

  4. 

  5. import scipy.stats as st
    6. 

  6. import sys
    7. 

  7. import matplotlib.pyplot as plt
    8. 

  8. import os
    9. 

  9. import seaborn as sns
    10. 

  10. from itertools import product
    11. 

  11. from collections import defaultdict
    12. 

  12. 

  13. 

  14. # dabele dla n1, n2 <20 http://www.real-statistics.com/statistics-tables/mann-whitney-table/
    15. 

  15. FILE_T = './U_statistic.csv'
    16. 

  16. FILE_S = './AUC_symulacje_full.csv'
    17. 

  17. FILE_S = './result.csv'
    18. 

  18. 

  19. min_per_round=2
    20. 

  20. 

  21. U_stat_diag = [    0.,     0.,     0.,     2.,     5.,     8.,    12.,    16.,
    22. 

  22.           22.,    28.,    35.,    43.,    52.,    62.,    73.,    84.,
    23. 

  23.           97.,   110.,   124.,   139.,   155.,   172.,   190.,   208.,
    24. 

  24.          228.,   248.,   269.,   292.,   315.,   339.,   364.,   389.,
    25. 

  25.          416.,   444.,   472.,   502.,   532.,   564.,   596.,   629.,
    26. 

  26.          663.,   698.,   734.,   771.,   809.,   847.,   887.,   927.,
    27. 

  27.          969.,  1011.,  1055.,  1099.,  1144.,  1190.,  1237.,  1285.,
    28. 

  28.         1334.,  1384.,  1435.,  1486.]
    29. 

  29. 

  30. 

  31. u_stat_01 = [1.0, 4.0, 7.0, 12.0, 19.0, 26.0, 35.0, 44.0, 55.0, 67.0, 80.0, 94.0, 110.0, 126.0, 144.0, 162.0, 182.0, 203.0, 225.0, 248.0, 272.0, 297.0, 323.0, 350.0, 379.0, 408.0, 439.0, 470.0, 503.0, 537.0, 572.0, 608.0, 645.0, 683.0, 722.0, 762.0, 803.0, 846.0, 889.0, 934.0, 979.0, 1026.0, 1073.0, 1122.0, 1172.0, 1222.0, 1274.0, 1327.0, 1381.0, 1436.0, 1492.0, 1550.0, 1608.0, 1667.0, 1727.0, 1789.0, 1851.0, 1915.0, 1979.0, 2045.0, 2111.0, 2179.0, 2248.0, 2317.0, 2388.0, 2460.0, 2533.0, 2607.0, 2682.0, 2758.0, 2835.0, 2913.0, 2992.0, 3073.0, 3154.0, 3236.0, 3320.0, 3404.0, 3490.0, 3576.0, 3664.0, 3752.0, 3842.0, 3932.0, 4024.0, 4117.0, 4211.0, 4306.0, 4402.0, 4498.0, 4596.0, 4695.0, 4796.0, 4897.0, 4999.0, 5102.0, 5206.0, 5311.0, 5418.0, 5525.0, 5633.0, 5743.0, 5853.0, 5965.0, 6077.0, 6191.0, 6305.0, 6421.0, 6538.0, 6656.0, 6774.0, 6894.0, 7015.0, 7137.0, 7260.0, 7384.0, 7509.0, 7635.0, 7762.0, 7890.0, 8019.0, 8149.0, 8280.0, 8412.0, 8546.0, 8680.0, 8815.0, 8952.0, 9089.0, 9228.0, 9367.0, 9508.0, 9649.0, 9792.0, 9935.0, 10080.0, 10226.0, 10372.0, 10520.0, 10669.0, 10819.0, 10970.0, 11121.0, 11274.0, 11428.0, 11583.0, 11739.0, 11896.0, 12054.0, 12213.0]
    32. 

  32. 

  33. u_stat_005 = [1.0, 4.0, 8.0, 14.0, 20.0, 28.0, 37.0, 48.0, 59.0, 72.0, 86.0, 101.0, 117.0, 134.0, 152.0, 172.0, 192.0, 214.0, 237.0, 261.0, 286.0, 312.0, 339.0, 368.0, 397.0, 428.0, 460.0, 492.0, 526.0, 561.0, 597.0, 635.0, 673.0, 712.0, 753.0, 794.0, 837.0, 880.0, 925.0, 971.0, 1018.0, 1066.0, 1115.0, 1165.0, 1216.0, 1269.0, 1322.0, 1377.0, 1432.0, 1489.0, 1546.0, 1605.0, 1665.0, 1726.0, 1788.0, 1851.0, 1915.0, 1980.0, 2046.0, 2114.0, 2182.0, 2251.0, 2322.0, 2393.0, 2466.0, 2540.0, 2614.0, 2690.0, 2767.0, 2845.0, 2924.0, 3004.0, 3085.0, 3167.0, 3250.0, 3335.0, 3420.0, 3506.0, 3594.0, 3682.0, 3772.0, 3862.0, 3954.0, 4047.0, 4140.0, 4235.0, 4331.0, 4428.0, 4526.0, 4625.0, 4725.0, 4826.0, 4929.0, 5032.0, 5136.0, 5241.0, 5348.0, 5455.0, 5564.0, 5673.0, 5784.0, 5896.0, 6008.0, 6122.0, 6237.0, 6353.0, 6470.0, 6588.0, 6707.0, 6827.0, 6948.0, 7070.0, 7193.0, 7317.0, 7443.0, 7569.0, 7696.0, 7825.0, 7954.0, 8085.0, 8216.0, 8349.0, 8483.0, 8617.0, 8753.0, 8890.0, 9028.0, 9167.0, 9306.0, 9447.0, 9589.0, 9732.0, 9877.0, 10022.0, 10168.0, 10315.0, 10463.0, 10613.0, 10763.0, 10915.0, 11067.0, 11220.0, 11375.0, 11531.0, 11687.0, 11845.0, 12004.0, 12163.0, 12324.0, 12486.0]
    34. 

  34. 

  35. u_stat_0025 = [1.0, 4.0, 9.0, 15.0, 22.0, 30.0, 40.0, 50.0, 63.0, 76.0, 90.0, 106.0, 123.0, 140.0, 160.0, 180.0, 201.0, 224.0, 247.0, 272.0, 298.0, 325.0, 353.0, 383.0, 413.0, 445.0, 478.0, 511.0, 546.0, 582.0, 619.0, 658.0, 697.0, 737.0, 779.0, 822.0, 865.0, 910.0, 956.0, 1003.0, 1052.0, 1101.0, 1151.0, 1203.0, 1255.0, 1309.0, 1363.0, 1419.0, 1476.0, 1534.0, 1593.0, 1653.0, 1714.0, 1777.0, 1840.0, 1904.0, 1970.0, 2037.0, 2104.0, 2173.0, 2243.0, 2314.0, 2386.0, 2459.0, 2533.0, 2608.0, 2685.0, 2762.0, 2840.0, 2920.0, 3000.0, 3082.0, 3165.0, 3249.0, 3334.0, 3419.0, 3506.0, 3595.0, 3684.0, 3774.0, 3865.0, 3957.0, 4051.0, 4145.0, 4241.0, 4338.0, 4435.0, 4534.0, 4634.0, 4735.0, 4837.0, 4940.0, 5044.0, 5149.0, 5255.0, 5362.0, 5470.0, 5580.0, 5690.0, 5802.0, 5914.0, 6028.0, 6142.0, 6258.0, 6375.0, 6493.0, 6612.0, 6732.0, 6853.0, 6975.0, 7098.0, 7222.0, 7347.0, 7474.0, 7601.0, 7729.0, 7859.0, 7989.0, 8121.0, 8253.0, 8387.0, 8522.0, 8658.0, 8795.0, 8932.0, 9071.0, 9211.0, 9352.0, 9495.0, 9638.0, 9782.0, 9927.0, 10074.0, 10221.0, 10369.0, 10519.0, 10669.0, 10821.0, 10974.0, 11127.0, 11282.0, 11438.0, 11595.0, 11752.0, 11911.0, 12071.0, 12232.0, 12394.0, 12558.0, 12722.0]
    36. 

  36. 

  37. 

  38. def AUC_T(U, N, p):
    39. 

  39.     # U wartsc statystyki
    40. 

  40.     # N liczba T i NT przed walidacja
    41. 

  41.     # p = p_value
    42. 

  42. 

  43.     R_out = []
    44. 

  44.     N_out = []
    45. 

  45. 

  46.     for u, n in zip(U, N):
    47. 

  47.         if np.isnan(u):
    48. 

  48.             R_out.append(np.nan)
    49. 

  49.             N_out.append(n)
    50. 

  50.         else:
    51. 

  51.             R_out.append((n**2 - u)/ n**2)
    52. 

  52.             N_out.append(n)
    53. 

  53. 

  54. 

  55. 

  56.     for n in range(20, 101):
    57. 

  57.         meanrank = n**2 / 2
    58. 

  58.         sd = np.sqrt(n**2 * (2*n+1) / 12.0)  # bez tiecorrection, ale powinno byc ok
    59. 

  59.         z = st.norm.ppf(1-p)  # test dwustronny
    60. 

  60.         u = z*sd + meanrank
    61. 

  61.         AUC = u / n**2
    62. 

  62.         R_out.append(AUC)
    63. 

  63.         N_out.append(n)
    64. 

  64.     return N_out, R_out
    65. 

  65. 

  66. 

  67. def main(file_T, file_S, file_DATABASE, NAME, main_diagnosis = None, auc_max=False):
    68. 

  68. 

  69.     AUC_T_data = pd.read_csv(file_T, sep=',')
    70. 

  70.     #p_data = AUC_T_data.columns.values[1:]
    71. 

  71.     
    72. 

  72.     p_data = ['0.05']
    73. 

  73.     
    74. 

  74.     print p_data
    75. 

  75. 

  76.     AUC_S_data = pd.read_csv(file_S, sep=',')
    77. 

  77.     AUC_DATABASE = pd.read_csv(file_DATABASE, sep=',')
    78. 

  78. 

  79.     f = plt.figure(figsize=(10,5), dpi=600)
    80. 

  80.     ax = f.add_subplot(1, 1, 1)
    81. 

  81.     color = {'0.01':'b', '0.05':'gray'}
    82. 

  82.     for ind, p in enumerate(p_data,1):
    83. 

  83. 

  84.         # f = plt.figure()
    85. 

  85.         # ax = f.add_subplot(1, 1, 1)
    86. 

  86. 

  87.         #AUC_TEORETYCZNE
    88. 

  88.         data_U = AUC_T_data[p].values
    89. 

  89.         data_n = AUC_T_data['n'].values
    90. 

  90.         N_values, AUC_values = AUC_T(data_U, data_n, float(p))
    91. 

  91.         ax.plot(N_values, AUC_values, color = color[p]  ,label = 'theoretical p={}'.format(p), alpha=0.75)
    92. 

  92.         
    93. 

  93.         #~ xx= []
    94. 

  94.         #~ yy =[]
    95. 

  95.         #~ for nr, i in enumerate(u_stat_01):
    96. 

  96.             #~ yy.append(i/(nr+1)**2)
    97. 

  97.             #~ xx.append(nr+1)
    98. 

  98.         #~ 
    99. 

  99.         #~ ax.plot(xx, yy, label='theory 0.1')
    100. 

  100.         #~ 
    101. 

  101.         #~ u_stat_005
    102. 

  102.         #~ 
    103. 

  103.         #~ xx= []
    104. 

  104.         #~ yy =[]
    105. 

  105.         #~ for nr, i in enumerate(u_stat_005):
    106. 

  106.             #~ yy.append(i/(nr+1)**2)
    107. 

  107.             #~ xx.append(nr+1)
    108. 

  108.         #~ 
    109. 

  109.         #~ ax.plot(xx, yy, label='theory 0.05')
    110. 

  110.         #~ 
    111. 

  111.         #~ xx= []
    112. 

  112.         #~ yy =[]
    113. 

  113.         #~ for nr, i in enumerate(u_stat_0025):
    114. 

  114.             #~ yy.append(i/(nr+1)**2)
    115. 

  115.             #~ xx.append(nr+1)
    116. 

  116.         #~ 
    117. 

  117.         #~ ax.plot(xx, yy, label='theory 0.025')
    118. 

  118. 

  119.         #AUC SYMULACJE
    120. 

  120. 

  121.         data_for_AUC = AUC_S_data[['n_targets', 'auc95' ]].sort_values('n_targets').values
    122. 

  122.         data_AUC = data_for_AUC[:,1]
    123. 

  123.         data_n = data_for_AUC[:,0]
    124. 

  124.         ax.plot(data_n, data_AUC, color = 'blue', label = 'simulation p={}'.format(p), alpha=0.75)
    125. 

  125.         
    126. 

  126.         #~ data_for_AUC = AUC_S_data[['n_targets', 'auc90' ]].sort_values('n_targets').values
    127. 

  127.         #~ data_AUC = data_for_AUC[:,1]
    128. 

  128.         #~ data_n = data_for_AUC[:,0]
    129. 

  129.         #~ ax.plot(data_n, data_AUC, color = 'red', label = 'AUC crit. simulation p={}'.format(0.1), alpha=0.75)
    130. 

  130. 

  131.     color = ['g', 'y', 'r', 'c', 'm', 'k']
    132. 

  132.     for i, diag in enumerate(sorted(list(set(AUC_DATABASE['diagnosis'].values)))):
    133. 

  133.         if auc_max:
    134. 

  134.             data_AUC_n, data_AUC = get_auc_max(AUC_DATABASE, diag)
    135. 

  135.         else:
    136. 

  136.             data_AUC = AUC_DATABASE[AUC_DATABASE['diagnosis']==diag]['AUC'].values
    137. 

  137.             data_AUC_n = AUC_DATABASE[AUC_DATABASE['diagnosis']==diag]['N_avr_targets'].values
    138. 

  138.         if main_diagnosis is None:
    139. 

  139.             color__ = color[i]
    140. 

  140.         else:
    141. 

  141.             if diag == main_diagnosis:
    142. 

  142.                 color__ = 'k'
    143. 

  143.             else:
    144. 

  144.                 color__ = 'lightgray'
    145. 

  145.             
    146. 

  146.         if diag=='CONTROL_ADULT':
    147. 

  147.             diag = 'control adult'
    148. 

  148.         if diag=='CONTROL_CHILDREN':
    149. 

  149.             diag = 'control children'
    150. 

  150.             
    151. 

  151.         # DATA FILTERING:
    152. 

  152.         mask = data_AUC_n>=20
    153. 

  153.         
    154. 

  154.         data_AUC_n = data_AUC_n[mask]
    155. 

  155.         data_AUC = data_AUC[mask]
    156. 

  156.         
    157. 

  157.         ax.plot(data_AUC_n, data_AUC, 'o', color = color__, label = '{}'.format(diag,), alpha=0.75)
    158. 

  158. 

  159.     ax.set_xlim(0, 100)
    160. 

  160.     ax.set_ylim(0.0, 1)
    161. 

  161.     ax.set_xlabel("N1 = N2 = N target and non-target example count")
    162. 

  162.     ax.set_ylabel("AUC")
    163. 

  163.     # plt.title("P = {} one-tailed".format(p))
    164. 

  164.     # plt.savefig('./AUC_{}.png'.format(p))
    165. 

  165. 

  166. 

  167. ############################
    168. 

  168.     if main_diagnosis:
    169. 

  169.         import matplotlib.lines as mlines
    170. 

  170.         theory_line = mlines.Line2D([], [], color='gray',
    171. 

  171.                                 label='theoretical p=0.05')
    172. 

  172.         sim_line = mlines.Line2D([], [], color='gray', marker='>', linestyle='',
    173. 

  173.                                 label='simulation p=0.05')
    174. 

  174.         
    175. 

  175.         main_diag_line = mlines.Line2D([], [], color='k', marker='o', linestyle='',
    176. 

  176.                                 label='{}'.format(main_diagnosis))
    177. 

  177.         
    178. 

  178.         other_diag_line = mlines.Line2D([], [], color='lightgray', marker='o', linestyle='',
    179. 

  179.                                 label='others')
    180. 

  180. 

  181.         plt.legend(handles=[theory_line, sim_line, main_diag_line, other_diag_line], loc=0)
    182. 

  182. ######################################################################
    183. 

  183.     else:
    184. 

  184.         plt.legend(loc=0)
    185. 

  185. 

  186.     plt.savefig('./AUC_{}_{}_max{}_min_per_round_{}.png'.format(main_diagnosis,NAME, auc_max, min_per_round), dpi=600)
    187. 

  187. 

  188. 

  189. 

  190. def get_auc_max(d, diag):
    191. 

  191.     d['crs_group'] = d.diagnosis
    192. 

  192.     # d.crs_group.loc[(d.crs_group=='MCS-') | (d.crs_group=='MCS+')] = 'MCS'
    193. 

  193. 

  194.     patients = d["id"].unique()
    195. 

  195.     d_max = pd.DataFrame(columns=d.columns)
    196. 

  196.     for p in patients:
    197. 

  197.             rounds = d["round"].loc[(d["id"]==p)].unique()
    198. 

  198.             for r in rounds:
    199. 

  199.                     p_vals = d["AUC"].loc[(d["id"]==p) & (d["round"]==r)]
    200. 

  200.                     temp = d.loc[p_vals.argmax()]
    201. 

  201.                     print('hh', r, p, len(p_vals), diag)
    202. 

  202.                     
    203. 

  203.                     to_append = temp.append(pd.Series({'nb_paradigms': len(p_vals)}))
    204. 

  204.                     if len(p_vals)>=min_per_round or diag=='UWS' or diag=='CONTROL_ADULT' or diag=='CONTROL_CHILDREN':
    205. 

  205.                         d_max = d_max.append(to_append, ignore_index=True)
    206. 

  206.                     
    207. 

  207.     data_AUC = d_max[d_max['diagnosis']==diag]['AUC'].values
    208. 

  208.     data_AUC_n = d_max[d_max['diagnosis']==diag]['N_avr_targets'].values
    209. 

  209.     
    210. 

  210.     return data_AUC_n, data_AUC
    211. 

  211. 

  212. 

  213. 

  214. 

  215. if __name__ == '__main__':
    216. 

  216.     database_AUC = [ ["AUC_downsamping_database_1.csv", "downsamping_1"],
    217. 

  217.             
    218. 

  218.             ]
    219. 

  219.     AUC_S_data = pd.read_csv(FILE_S, sep=',')
    220. 

  220.     AUC_DATABASE = pd.read_csv('AUC_downsamping_database_1.csv', sep=',')
    221. 

  221.     MAIN_DIAGNOSES = set(AUC_DATABASE['diagnosis'].values)
    222. 

  222.     for file_, name in database_AUC:
    223. 

  223.         data = main(FILE_T, FILE_S, file_, name, None, True)
    224. 

  224.         data = main(FILE_T, FILE_S, file_, name, None, False)
    225. 

  225.     #for main_diagnosis in MAIN_DIAGNOSES:
    226. 

  226.     #       for file_, name in database_AUC: 
    227. 

  227.     #        
    228. 

  228.     #        data = main(FILE_T, FILE_S, file_, name, main_diagnosis, True)
    229. 

  229.