/
plots_experimental_data.py
1486 lines (1175 loc) · 69 KB
/
plots_experimental_data.py
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
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import sys
import numpy as np
import scipy.io as sio
import matplotlib.pyplot as plt
# import matplotlib.patches as plt_patches
# import matplotlib.gridspec as plt_grid
import os
import os.path
import cPickle as pickle
# import bottleneck as bn
import em_circularmixture
import em_circularmixture_allitems_uniquekappa
import pandas as pd
import matplotlib.cm as cmx
import matplotlib.colors as mcolors
from mpl_toolkits.mplot3d import Axes3D
import seaborn as sns
import dataio as DataIO
import utils
def plot_kappa_mean_error(T_space, mean, yerror, ax=None, dataio=None, title='',
fmt='o-', legend=True, **args):
'''
!!! IMPORTANT !!!
Main plotting function to show the evolution of Kappa with T
!!! IMPORTANT !!!
'''
if ax is None:
f, ax = plt.subplots()
ax = utils.plot_mean_std_area(
T_space,
mean,
np.ma.masked_invalid(yerror).filled(0.0),
ax_handle=ax,
linewidth=3,
markersize=8,
fmt=fmt,
**args)
if legend:
ax.legend(prop={'size': 15}, loc='best')
if title:
ax.set_title('Kappa: %s' % title)
ax.set_xlim([0.9, T_space.max()+0.1])
ax.set_ylim([0.0, max(np.max(mean)*1.1, ax.get_ylim()[1])])
ax.set_xticks(range(1, T_space.max()+1))
ax.set_xticklabels(range(1, T_space.max()+1))
ax.get_figure().canvas.draw()
if dataio is not None:
dataio.save_current_figure('%s_kappa_{label}_{unique_id}.pdf' % title)
return ax
def plot_emmixture_mean_error(T_space, mean, yerror, ax=None, dataio=None,
title='', fmt='o-', legend=True, **args):
'''
!!! IMPORTANT !!!
Main plotting function to show the evolution of an EM Mixture proportion with T
!!! IMPORTANT !!!
'''
if ax is None:
f, ax = plt.subplots()
utils.plot_mean_std_area(
T_space, mean, np.ma.masked_invalid(yerror).filled(0.0),
ax_handle=ax, linewidth=3, fmt=fmt, markersize=8, **args)
if legend:
ax.legend(prop={'size': 15}, loc='best')
if title:
ax.set_title('Mixture prop: %s' % title)
ax.set_xlim([0.9, T_space.max() + 0.1])
ax.set_ylim([0.0, 1.01])
ax.set_xticks(range(1, T_space.max()+1))
ax.set_xticklabels(range(1, T_space.max()+1))
ax.get_figure().canvas.draw()
if dataio is not None:
dataio.save_current_figure('%s_mixtures_{label}_{unique_id}.pdf' % title)
return ax
def plot_check_bias_nontarget(dataset, dataio=None):
'''
Get an histogram of the errors between the response and all non targets
If biased towards 0-values, should indicate misbinding errors.
(if you do this with respect to all targets, it's retarded and should always be biased)
'''
n_items_space = np.unique(dataset['n_items'])
angle_space = np.linspace(-np.pi, np.pi, 20)
# Get histograms of errors, per n_item
for nitems_i in xrange(n_items_space.size):
utils.hist_samples_density_estimation(dataset['errors_nitems'][nitems_i], bins=angle_space, title='%s N=%d' % (dataset['name'], n_items_space[nitems_i]), dataio=dataio, filename='hist_bias_targets_%ditems_{label}_{unique_id}.pdf' % (n_items_space[nitems_i]))
# Get histograms of bias to nontargets. Do that by binning the errors to others nontargets of the array.
utils.plot_hists_bias_nontargets(dataset['errors_all_nitems'][n_items_space>1], bins=20, dataio=dataio, label='allnontargets', remove_first_column=True)
rayleigh = utils.rayleigh(dataset['errors_all_nitems'][n_items_space>1].flatten())
v_test = utils.V_test(dataset['errors_all_nitems'][n_items_space>1].flatten())
print rayleigh
print v_test
def plot_check_bias_bestnontarget(dataset, dataio=None):
'''
Get an histogram of errors between response and best nontarget.
Should be more biased towards 0 than the overall average
'''
n_items_space = np.unique(dataset['n_items'])
# Compute the errors to the best non target
errors_nontargets = dataset['errors_all_nitems'][n_items_space>1]
errors_nontargets = np.array([errors_nontargets_nitem[~np.all(np.isnan(errors_nontargets_nitem), axis=1), :] for errors_nontargets_nitem in errors_nontargets])
indices_bestnontarget = [np.nanargmin(np.abs(errors_nontargets[n_item_i][..., 1:]), axis=-1) for n_item_i in xrange(errors_nontargets.shape[0])]
# indices_bestnontarget = np.nanargmin(np.abs(errors_nontargets), axis=2)
# Index of the argmin of absolute error. Not too bad, easy to index into.
errors_bestnontargets_nitems = np.array([ errors_nontargets[n_items_i][ xrange(errors_nontargets[n_items_i].shape[0]), indices_bestnontarget[n_items_i] + 1] for n_items_i in xrange(errors_nontargets.shape[0]) ])
# Show histograms per n_items, like in Bays2009 figure
utils.plot_hists_bias_nontargets(errors_bestnontargets_nitems, bins=20, label='bestnontarget', dataio=dataio)
def plot_check_bias_nontarget_randomized(dataset, dataio=None):
'''
Plot the histogram of errors to nontargets, after replacing all nontargets by random angles.
If show similar bias, would be indication of low predictive power of distribution of errors to nontargets.
'''
n_items_space = np.unique(dataset['n_items'])
# Copy item_angles
new_item_angles = dataset['item_angle'].copy()
# Will resample multiple times
errors_nitems_new_dict = dict()
nb_resampling = 100
for resampling_i in xrange(nb_resampling):
# Replace nontargets randomly
nontarget_indices = np.nonzero(~np.isnan(new_item_angles[:, 1:]))
new_item_angles[nontarget_indices[0], nontarget_indices[1]+1] = 2*np.pi*np.random.random(nontarget_indices[0].size) - np.pi
# Compute errors
new_all_errors = utils.wrap_angles(new_item_angles - dataset['response'], bound=np.pi)
for n_items in n_items_space:
ids_filtered = (dataset['n_items'] == n_items).flatten()
if n_items in errors_nitems_new_dict:
errors_nitems_new_dict[n_items] = np.r_[errors_nitems_new_dict[n_items], new_all_errors[ids_filtered]]
else:
errors_nitems_new_dict[n_items] = new_all_errors[ids_filtered]
errors_nitems_new = np.array([val for key, val in errors_nitems_new_dict.items()])
utils.plot_hists_bias_nontargets(errors_nitems_new[n_items_space>1], bins=20, label='allnontarget_randomized_%dresamplings' % nb_resampling, dataio=dataio, remove_first_column=True)
### Do same for best non targets
# TODO Convert this for data_dualrecall
errors_nontargets = errors_nitems_new[1:, :, 1:]
indices_bestnontarget = np.nanargmin(np.abs(errors_nontargets), axis=2)
# Index of the argmin of absolute error. Not too bad, easy to index into.
errors_bestnontargets_nitems = np.array([ errors_nontargets[n_items_i, xrange(errors_nontargets.shape[1]), indices_bestnontarget[n_items_i]] for n_items_i in xrange(errors_nontargets.shape[0]) ])
# Show histograms
utils.plot_hists_bias_nontargets(errors_bestnontargets_nitems, bins=20, label='bestnontarget_randomized_%dresamplings' % nb_resampling, dataio=dataio)
def plot_check_oblique_effect(data, nb_bins=100):
'''
Humans are more precise for vertical and horizontal bars than diagonal orientations.
Check if present.
'''
# Construct the list of (target angles, errors), see if there is some structure in that
errors_per_angle = np.array(zip(data['item_angle'][np.arange(data['probe'].size), data['probe'][:, 0]], data['error'][:, 0]))
# response_per_angle = np.array(zip(data['item_angle'][np.arange(data['probe'].size), data['probe'][:, 0]], data['response']))
# response_per_colour = np.array(zip(data['item_colour'][np.arange(data['probe'].size), data['probe'][:, 0]], data['response']))
plt.figure()
plt.plot(errors_per_angle[:, 0], errors_per_angle[:, 1], 'x')
plt.figure()
plt.plot(errors_per_angle[:, 0], np.abs(errors_per_angle[:, 1]), 'x')
discrete_x = np.linspace(-np.pi/2., np.pi/2., nb_bins)
avg_error = np.zeros(discrete_x.shape)
std_error = np.zeros(discrete_x.shape)
for x_i in np.arange(discrete_x.size):
if x_i < discrete_x.size - 1:
# Check what data comes in the current interval x[x_i, x_i+1]
avg_error[x_i] = utils.mean_angles(errors_per_angle[np.logical_and(errors_per_angle[:, 0] > discrete_x[x_i], errors_per_angle[:, 0] < discrete_x[x_i+1]), 1])
std_error[x_i] = utils.angle_circular_std_dev(errors_per_angle[np.logical_and(errors_per_angle[:, 0] > discrete_x[x_i], errors_per_angle[:, 0] < discrete_x[x_i+1]), 1])
plt.figure()
plt.plot(discrete_x, avg_error)
plt.figure()
plt.plot(discrete_x, avg_error**2.)
plt.figure()
plt.plot(discrete_x, np.abs(avg_error))
plt.figure()
plt.plot(errors_per_angle[:, 0], errors_per_angle[:, 1], 'x')
plt.plot(discrete_x, avg_error, 'ro')
def plot_histograms_errors_targets_nontargets_nitems(dataset, dataio=None):
'''
Create subplots showing histograms of errors to targets and nontargets
Adds Vtest texts on the nontargets
'''
angle_space = np.linspace(-np.pi, np.pi, 51)
bins_center = angle_space[:-1] + np.diff(angle_space)[0]/2
#### Histograms of errors / errors to nontargets, collapsing across subjects
if True:
f1, axes1 = plt.subplots(ncols=dataset['n_items_size'],
nrows=2,
figsize=(dataset['n_items_size']*6, 12),
)
for n_items_i, n_items in enumerate(np.unique(dataset['n_items'])):
utils.hist_angular_data(
dataset['errors_nitems'][n_items_i],
bins=angle_space,
title='N=%d' % (n_items),
norm='density',
ax_handle=axes1[0, n_items_i],
pretty_xticks=True)
axes1[0, n_items_i].set_ylim([0., 2.0])
if n_items > 1:
utils.hist_angular_data(
utils.dropnan(dataset['errors_nontarget_nitems'][n_items_i]),
bins=angle_space,
# title='Nontarget %s N=%d' % (dataset['name'], n_items),
norm='density',
ax_handle=axes1[1, n_items_i],
pretty_xticks=True)
axes1[1, n_items_i].text(
0.02,
0.96,
"Vtest pval: %.4f" % (dataset['vtest_nitems'][n_items_i]),
transform=axes1[1, n_items_i].transAxes,
horizontalalignment='left',
fontsize=13)
axes1[1, n_items_i].set_ylim([0., 0.3])
else:
axes1[1, n_items_i].axis('off')
# f1.suptitle(dataset['name'])
f1.canvas.draw()
if dataio is not None:
plt.figure(f1.number)
# plt.tight_layout()
dataio.save_current_figure("hist_error_all_{label}_{unique_id}.pdf")
### Histograms per subject and nitems
if True:
f2, axes2 = plt.subplots(ncols=dataset['n_items_size'],
nrows=2,
figsize=(dataset['n_items_size']*6, 12))
for n_items_i, n_items in enumerate(np.unique(dataset['n_items'])):
axes2[0, n_items_i].bar(
bins_center,
dataset['hist_cnts_target_nitems_stats']['mean'][n_items_i],
width=2.*np.pi/(angle_space.size-1),
align='center',
yerr=dataset['hist_cnts_target_nitems_stats']['sem'][n_items_i])
# axes3[n_items_i].set_title('N=%d' % n_items)
axes2[0, n_items_i].set_title('N=%d' % n_items)
axes2[0, n_items_i].set_xlim(
[bins_center[0]-np.pi/(angle_space.size-1),
bins_center[-1]+np.pi/(angle_space.size-1)])
axes2[0, n_items_i].set_ylim([0., 2.0])
axes2[0, n_items_i].set_xticks((-np.pi, -np.pi/2, 0, np.pi/2., np.pi))
axes2[0, n_items_i].set_xticklabels(
(r'$-\pi$',
r'$-\frac{\pi}{2}$',
r'$0$',
r'$\frac{\pi}{2}$',
r'$\pi$'),
fontsize=16)
if n_items > 1:
axes2[1, n_items_i].bar(
bins_center,
dataset['hist_cnts_nontarget_nitems_stats']['mean'][n_items_i],
yerr=dataset['hist_cnts_nontarget_nitems_stats']['sem'][n_items_i],
width=2.*np.pi/(angle_space.size-1),
align='center')
# axes2[1, n_items_i].set_title('N=%d' % n_items)
# axes2[1, n_items_i].text(0.02, 0.96, "Vtest pval: %.4f" % (pvalue_nontarget_subject_nitems_mean[n_items_i]), transform=axes2[1, n_items_i].transAxes, horizontalalignment='left', fontsize=13)
axes2[1, n_items_i].text(
0.02,
0.96,
"Vtest pval: %.4f" % (dataset['vtest_nitems'][n_items_i]),
transform=axes2[1, n_items_i].transAxes,
horizontalalignment='left',
fontsize=14)
# TODO Add bootstrap there instead.
# axes2[1, n_items_i].set_title('Nontarget, subject, N=%d' % n_items)
axes2[1, n_items_i].set_xlim(
[bins_center[0]-np.pi/(angle_space.size-1),
bins_center[-1]+np.pi/(angle_space.size-1)])
axes2[1, n_items_i].set_ylim([0., 0.3])
axes2[1, n_items_i].set_xticks(
(-np.pi, -np.pi/2, 0, np.pi/2., np.pi))
axes2[1, n_items_i].set_xticklabels(
(r'$-\pi$',
r'$-\frac{\pi}{2}$',
r'$0$',
r'$\frac{\pi}{2}$',
r'$\pi$'),
fontsize=16)
else:
axes2[1, n_items_i].axis('off')
# f2.suptitle(dataset['name'])
f2.canvas.draw()
if dataio is not None:
plt.figure(f2.number)
# plt.tight_layout()
dataio.save_current_figure("hist_error_persubj_{label}_{unique_id}.pdf")
### Scatter marginals, 3-parts figures
if True:
f3_all = []
for n_items_i, n_items in enumerate(np.unique(dataset['n_items'])):
f3 = utils.scatter_marginals(
utils.dropnan(
dataset['data_to_fit'][n_items]['item_features'][:, 0, 0]),
utils.dropnan(dataset['data_to_fit'][n_items]['response']),
xlabel='Target',
ylabel='Response',
# title='%s, %d items' % (
# dataset['name'], n_items),
figsize=(9, 9),
factor_axis=1.1,
bins=61)
f3_all.append(f3)
return f1, f2, f3_all
def plot_em_mixtures(dataset, dataio=None, use_sem=True):
'''
Do plots for the mixture models and kappa
'''
T_space_exp = np.unique(dataset['n_items'])
f, ax = plt.subplots()
if use_sem:
errorbars = 'sem'
else:
errorbars = 'std'
# Mixture probabilities
utils.plot_mean_std_area(T_space_exp, dataset['em_fits_nitems_arrays']['mean'][1], np.ma.masked_invalid(dataset['em_fits_nitems_arrays'][errorbars][1]).filled(0.0), xlabel='Number of items', ylabel="Mixture probabilities", ax_handle=ax, linewidth=3, fmt='o-', markersize=5, label='Target')
utils.plot_mean_std_area(T_space_exp, np.ma.masked_invalid(dataset['em_fits_nitems_arrays']['mean'][2]).filled(0.0), np.ma.masked_invalid(dataset['em_fits_nitems_arrays'][errorbars][2]).filled(0.0), xlabel='Number of items', ylabel="Mixture probabilities", ax_handle=ax, linewidth=3, fmt='o-', markersize=5, label='Nontarget')
utils.plot_mean_std_area(T_space_exp, dataset['em_fits_nitems_arrays']['mean'][3], np.ma.masked_invalid(dataset['em_fits_nitems_arrays'][errorbars][3]).filled(0.0), xlabel='Number of items', ylabel="Mixture probabilities", ax_handle=ax, linewidth=3, fmt='o-', markersize=5, label='Random')
ax.legend(prop={'size':15})
ax.set_title('Mixture model for EM fit %s' % dataset['name'])
ax.set_xlim([1.0, T_space_exp.max()])
ax.set_ylim([0.0, 1.1])
ax.set_xticks(range(1, T_space_exp.max()+1))
ax.set_xticklabels(range(1, T_space_exp.max()+1))
f.canvas.draw()
if dataio is not None:
dataio.save_current_figure('emfits_mixtures_{label}_{unique_id}.pdf')
# Kappa
f, ax = plt.subplots()
ax = utils.plot_mean_std_area(T_space_exp,
dataset['em_fits_nitems_arrays']['mean'][0], np.ma.masked_invalid(dataset['em_fits_nitems_arrays'][errorbars][0]).filled(0.0), linewidth=3, fmt='o-', markersize=8, label='Kappa', xlabel='Number of items', ylabel='Experimental data', ax_handle=ax)
ax.legend(prop={'size':15})
ax.set_title('Kappa for EM fit %s' % dataset['name'])
ax.set_xlim([0.9, T_space_exp.max()+0.1])
ax.set_ylim([0.0, np.max(dataset['em_fits_nitems_arrays']['mean'][0])*1.1])
ax.set_xticks(range(1, T_space_exp.max()+1))
ax.set_xticklabels(range(1, T_space_exp.max()+1))
ax.get_figure().canvas.draw()
if dataio is not None:
dataio.save_current_figure('emfits_kappa_{label}_{unique_id}.pdf')
def plot_precision(dataset, dataio=None, use_sem=True):
'''
Do plots for the mixture models and kappa
'''
T_space_exp = np.unique(dataset['n_items'])
precisions_to_plot = [['precision_subject_nitems_theo', 'Precision Theo'],['precision_subject_nitems_bays_notreatment', 'Precision BaysNoTreat'],['precision_subject_nitems_bays', 'Precision Bays'],['precision_subject_nitems_theo_nochance', 'Precision TheoNoChance']]
for precision_to_plot, precision_title in precisions_to_plot:
f, ax = plt.subplots()
# Compute the errorbars
precision_mean = np.mean(dataset[precision_to_plot], axis=0)
precision_errors = np.std(dataset[precision_to_plot], axis=0)
if use_sem:
precision_errors /= np.sqrt(dataset['subject_size'])
# Now show the precision
utils.plot_mean_std_area(T_space_exp, precision_mean, precision_errors, xlabel='Number of items', label="Precision", ax_handle=ax, linewidth=3, fmt='o-', markersize=5)
ax.legend(prop={'size':15})
ax.set_title('%s %s' % (precision_title, dataset['name']))
ax.set_xlim([1.0, T_space_exp.max()])
ax.set_ylim([0.0, np.max(precision_mean)+np.max(precision_errors)])
ax.set_xticks(range(1, T_space_exp.max()+1))
ax.set_xticklabels(range(1, T_space_exp.max()+1))
f.canvas.draw()
if dataio is not None:
dataio.save_current_figure('%s_{label}_{unique_id}.pdf' % precision_title)
def plot_dualrecall(dataset):
'''
Create plots for the double recall dataset
'''
to_plot = {'resp_vs_targ':True, 'error_boxplot':True, 'resp_rating':True, 'em_fits':True, 'loglik':True, 'resp_distrib':True, 'resp_conds':True}
dataset_pd = dataset['panda']
dataset_pd['error_abs'] = dataset_pd.error.abs()
# Show distributions of responses wrt target angles/colour
if to_plot['resp_vs_targ']:
# Plot scatter and marginals for the orientation trials
utils.scatter_marginals(utils.dropnan(dataset['item_angle'][dataset['angle_trials'] & dataset['3_items_trials'], 0]), utils.dropnan(dataset['probe_angle'][dataset['angle_trials'] & dataset['3_items_trials']]), xlabel ='Target angle', ylabel='Response angle', title='%s Angle trials, 3 items' % (dataset['name']), figsize=(9, 9), factor_axis=1.1, bins=61)
utils.scatter_marginals(utils.dropnan(dataset['item_angle'][dataset['angle_trials'] & dataset['6_items_trials'], 0]), utils.dropnan(dataset['probe_angle'][dataset['angle_trials'] & dataset['6_items_trials']]), xlabel ='Target angle', ylabel='Response angle', title='%s Angle trials, 6 items' % (dataset['name']), figsize=(9, 9), factor_axis=1.1, bins=61)
# Plot scatter and marginals for the colour trials
utils.scatter_marginals(utils.dropnan(dataset['item_colour'][dataset['colour_trials']& dataset['3_items_trials'], 0]), utils.dropnan(dataset['probe_colour'][dataset['colour_trials'] & dataset['3_items_trials']]), xlabel ='Target colour', ylabel='Response colour', title='%s Colour trials, 3 items' % (dataset['name']), figsize=(9, 9), factor_axis=1.1, bins=61, show_colours=True)
utils.scatter_marginals(utils.dropnan(dataset['item_colour'][dataset['colour_trials'] & dataset['6_items_trials'], 0]), utils.dropnan(dataset['probe_colour'][dataset['colour_trials'] & dataset['6_items_trials']]), xlabel ='Target colour', ylabel='Response colour', title='%s Colour trials, 6 items' % (dataset['name']), figsize=(9, 9), factor_axis=1.1, bins=61, show_colours=True)
if 'em_fits' in dataset:
# dataset_pd[ids_filtered][ids_targets_responses].boxplot('error_angle_abs', by='rating')
# dataset_pd[ids_filtered][ids_nontargets_responses].boxplot('error_angle_abs', by='rating')
if to_plot['error_boxplot']:
dataset_pd.boxplot(column=['error_abs'], by=['cond', 'n_items', 'rating'])
# for i in dataset_pd.subject.unique():
# dataset_pd[dataset_pd.subject == i].boxplot(column=['error_angle'], by=['n_items', 'rating'])
# Show distribution responsibility as a function of rating
if to_plot['resp_rating']:
# dataset_grouped_nona_rating = dataset_pd[dataset_pd.n_items == 3.0].dropna(subset=['error']).groupby(['rating'])
dataset_grouped_nona_rating = dataset_pd[dataset_pd.n_items == 6.0][dataset_pd.cond == 1.].dropna(subset=['error']).groupby(['rating'])
_, axes = plt.subplots(dataset_pd.rating.nunique(), 3)
i = 0
bins = np.linspace(0., 1.0, 31)
for name, group in dataset_grouped_nona_rating:
print name
# Compute histograms and normalize per rating condition
counts_target, bins_edges = np.histogram(group.resp_target, bins=bins)
counts_nontarget, bins_edges = np.histogram(group.resp_nontarget, bins=bins)
counts_random, bins_edges = np.histogram(group.resp_random, bins=bins)
dedges = np.diff(bins_edges)[0]
sum_counts = float(np.sum(counts_target) + np.sum(counts_nontarget) + np.sum(counts_random))
counts_target = counts_target/sum_counts
counts_nontarget = counts_nontarget/sum_counts
counts_random = counts_random/sum_counts
# Print Responsibility target density estimation
# group.resp_target.plot(kind='kde', ax=axes[i, 0])
axes[i, 0].bar(bins_edges[:-1], counts_target, dedges, color='b')
axes[i, 0].set_xlim((0.0, 1.0))
axes[i, 0].set_ylim((0.0, 0.35))
axes[i, 0].text(0.5, 0.8, "T " + str(name), fontweight='bold', horizontalalignment='center', transform = axes[i, 0].transAxes)
# Print Responsibility nontarget density estimation
# group.resp_nontarget.plot(kind='kde', ax=axes[i, 1])
axes[i, 1].bar(bins_edges[:-1], counts_nontarget, dedges, color='g')
axes[i, 1].set_xlim((0.0, 1.0))
axes[i, 1].set_ylim((0.0, 0.35))
axes[i, 1].text(0.5, 0.8, "NT " + str(name), fontweight='bold', horizontalalignment='center', transform = axes[i, 1].transAxes)
# Print Responsibility random density estimation
# group.resp_random.plot(kind='kde', ax=axes[i, 1])
axes[i, 2].bar(bins_edges[:-1], counts_random, dedges, color='r')
axes[i, 2].set_xlim((0.0, 1.0))
axes[i, 2].set_ylim((0.0, 0.35))
axes[i, 2].text(0.5, 0.8, "R " + str(name), fontweight='bold', horizontalalignment='center', transform = axes[i, 2].transAxes)
i += 1
plt.suptitle("Colour trials")
dataset_grouped_nona_rating = dataset_pd[dataset_pd.n_items == 6.0][dataset_pd.cond == 2.].dropna(subset=['error']).groupby(['rating'])
f, axes = plt.subplots(dataset_pd.rating.nunique(), 3)
i = 0
bins = np.linspace(0., 1.0, 31)
for name, group in dataset_grouped_nona_rating:
print name
# Compute histograms and normalize per rating condition
counts_target, bins_edges = np.histogram(group.resp_target, bins=bins)
counts_nontarget, bins_edges = np.histogram(group.resp_nontarget, bins=bins)
counts_random, bins_edges = np.histogram(group.resp_random, bins=bins)
dedges = np.diff(bins_edges)[0]
sum_counts = float(np.sum(counts_target) + np.sum(counts_nontarget) + np.sum(counts_random))
counts_target = counts_target/sum_counts
counts_nontarget = counts_nontarget/sum_counts
counts_random = counts_random/sum_counts
# Print Responsibility target density estimation
# group.resp_target.plot(kind='kde', ax=axes[i, 0])
axes[i, 0].bar(bins_edges[:-1], counts_target, dedges, color='b')
axes[i, 0].set_xlim((0.0, 1.0))
axes[i, 0].set_ylim((0.0, 0.35))
axes[i, 0].text(0.5, 0.8, "T " + str(name), fontweight='bold', horizontalalignment='center', transform = axes[i, 0].transAxes)
# Print Responsibility nontarget density estimation
# group.resp_nontarget.plot(kind='kde', ax=axes[i, 1])
axes[i, 1].bar(bins_edges[:-1], counts_nontarget, dedges, color='g')
axes[i, 1].set_xlim((0.0, 1.0))
axes[i, 1].set_ylim((0.0, 0.35))
axes[i, 1].text(0.5, 0.8, "NT " + str(name), fontweight='bold', horizontalalignment='center', transform = axes[i, 1].transAxes)
# Print Responsibility random density estimation
# group.resp_random.plot(kind='kde', ax=axes[i, 1])
axes[i, 2].bar(bins_edges[:-1], counts_random, dedges, color='r')
axes[i, 2].set_xlim((0.0, 1.0))
axes[i, 2].set_ylim((0.0, 0.35))
axes[i, 2].text(0.5, 0.8, "R " + str(name), fontweight='bold', horizontalalignment='center', transform = axes[i, 2].transAxes)
i += 1
plt.suptitle("Angle trials")
# Add condition names
dataset_pd['cond_name'] = np.array(['Colour', 'Angle'])[np.array(dataset_pd['cond']-1, dtype=int)]
# Regroup some data
dataset_grouped_nona_conditems = dataset_pd.dropna(subset=['error']).groupby(['cond_name', 'n_items'])
dataset_grouped_nona_conditems_mean = dataset_grouped_nona_conditems.mean()[['mixt_target', 'mixt_nontargets_sum', 'mixt_random', 'kappa', 'train_LL', 'test_LL']]
# Show inferred mixture proportions and kappa
if to_plot['em_fits']:
# ax = dataset_grouped_nona_conditems_mean[['mixt_target', 'mixt_nontargets_sum', 'mixt_random', 'kappa']].plot(secondary_y='kappa', kind='bar')
ax = dataset_grouped_nona_conditems_mean[['mixt_target', 'mixt_nontargets_sum', 'mixt_random']].plot(kind='bar')
ax.set_ylabel('Mixture proportions')
ax = dataset_grouped_nona_conditems_mean[['kappa']].plot(kind='bar')
ax.set_ylabel('Kappa')
# Show loglihood of fit
if to_plot['loglik']:
f, ax = plt.subplots(1, 1)
dataset_grouped_nona_conditems_mean[['train_LL', 'test_LL']].plot(kind='bar', ax=ax, secondary_y='test_LL')
# Show boxplot of responsibilities
if to_plot['resp_distrib']:
dataset_grouped_nona_conditems.boxplot(column=['resp_target', 'resp_nontarget', 'resp_random'])
# Show distributions of responsibilities
if to_plot['resp_conds']:
f, axes = plt.subplots(dataset_pd.cond_name.nunique()*dataset_pd.n_items.nunique(), 3)
i = 0
bins = np.linspace(0., 1.0, 31)
for name, group in dataset_grouped_nona_conditems:
print name
# Print Responsibility target density estimation
# group.resp_target.plot(kind='kde', ax=axes[i, 0])
group.resp_target.hist(ax=axes[i, 0], color='b', bins=bins)
axes[i, 0].text(0.5, 0.85, "T " + ' '.join([str(x) for x in name]), fontweight='bold', horizontalalignment='center', transform = axes[i, 0].transAxes)
axes[i, 0].set_xlim((0.0, 1.0))
# Print Responsibility nontarget density estimation
# group.resp_nontarget.plot(kind='kde', ax=axes[i, 1])
group.resp_nontarget.hist(ax=axes[i, 1], color='g', bins=bins)
axes[i, 1].text(0.5, 0.85, "NT " + ' '.join([str(x) for x in name]), fontweight='bold', horizontalalignment='center', transform = axes[i, 1].transAxes)
axes[i, 1].set_xlim((0.0, 1.0))
# Print Responsibility random density estimation
# group.resp_random.plot(kind='kde', ax=axes[i, 1])
group.resp_random.hist(ax=axes[i, 2], color='r', bins=bins)
axes[i, 2].text(0.5, 0.85, "R " + ' '.join([str(x) for x in name]), fontweight='bold', horizontalalignment='center', transform = axes[i, 2].transAxes)
axes[i, 2].set_xlim((0.0, 1.0))
i += 1
# Extract some parameters
fitted_parameters = dataset_grouped_nona_conditems_mean.iloc[0].loc[['kappa', 'mixt_target', 'mixt_nontargets_sum', 'mixt_random']]
print fitted_parameters
def plot_bias_close_feature(dataset, dataio=None, markersize=3):
'''
Check if there is a bias in the response towards closest item (either closest wrt cued feature, or wrt all features)
'''
number_items_considered = 2
# Error to nontarget
bias_to_nontarget = np.abs(dataset['errors_nontarget_nitems'][number_items_considered-1][:, :number_items_considered-1].flatten())
bias_to_target = np.abs(dataset['errors_nitems'][number_items_considered-1].flatten())
ratio_biases = bias_to_nontarget/ bias_to_target
response = dataset['data_to_fit'][number_items_considered]['response']
target = dataset['data_to_fit'][number_items_considered]['item_features'][:, 0]
nontarget = dataset['data_to_fit'][number_items_considered]['item_features'][:, 1]
# Distance between probe and closest nontarget, in full feature space
dist_target_nontarget_torus = utils.dist_torus(target, nontarget)
# Distance only looking at recalled feature
dist_target_nontarget_recalled = np.abs(utils.wrap_angles((target[:, 0] - nontarget[:, 0])))
# Distance only looking at cued feature.
# Needs more work. They are only a few possible values, so we can group them and get a boxplot for each
dist_target_nontarget_cue = np.round(np.abs(utils.wrap_angles((target[:, 1] - nontarget[:, 1]))), decimals=8)
dist_distinct_values = np.unique(dist_target_nontarget_cue)
bias_to_nontarget_grouped_dist_cue = []
for dist_value in dist_distinct_values:
bias_to_nontarget_grouped_dist_cue.append(bias_to_nontarget[dist_target_nontarget_cue == dist_value])
# Check if the response is closer to the target or nontarget, in relative terms.
# Need to compute a ratio linking bias_to_target and bias_to_nontarget.
# Two possibilities: response was between target and nontarget, or response was "behind" the target.
ratio_response_close_to_nontarget = bias_to_nontarget/dist_target_nontarget_recalled
indices_filter_other_side = bias_to_nontarget > dist_target_nontarget_recalled
ratio_response_close_to_nontarget[indices_filter_other_side] = bias_to_nontarget[indices_filter_other_side]/(dist_target_nontarget_recalled[indices_filter_other_side] + bias_to_target[indices_filter_other_side])
f, ax = plt.subplots(2, 2, figsize=(8, 8))
ax[0, 0].plot(
dist_target_nontarget_torus, bias_to_nontarget, 'o',
markersize=markersize)
ax[0, 0].set_xlabel('Distance full feature space')
ax[0, 0].set_ylabel('Error to nontarget')
ax[0, 0].set_xlim((0, 4.5))
ax[0, 0].set_ylim((0, np.pi))
sns.boxplot(
x=dist_distinct_values, y=bias_to_nontarget_grouped_dist_cue,
orient='v', ax=ax[0, 1])
# ax[0, 1].boxplot(bias_to_nontarget_grouped_dist_cue, positions=dist_distinct_values)
ax[0, 1].set_ylabel('Error to nontarget')
ax[0, 1].set_xlabel('Distance cued feature only')
ax[0, 1].set_xticklabels(["%.2f" % x for x in dist_distinct_values])
ax[0, 1].set_ylim((0, np.pi))
ax[1, 0].plot(dist_target_nontarget_recalled, bias_to_nontarget, 'o',
markersize=markersize)
ax[1, 0].set_xlabel('Distance recalled feature only')
ax[1, 0].set_ylabel('Error to nontarget')
ax[1, 0].set_xlim((0, np.pi))
ax[1, 0].set_ylim((0, np.pi))
ax[1, 1].plot(dist_target_nontarget_recalled,
ratio_response_close_to_nontarget,
'o', markersize=markersize)
ax[1, 1].set_xlabel('Distance recalled feature only')
ax[1, 1].set_ylabel('Normalised distance\nto nontarget')
ax[1, 1].set_ylim((0, 1.05))
ax[1, 1].set_xlim((0, np.pi))
f.tight_layout()
if dataio:
f.set_size_inches(16, 16, forward=True)
dataio.save_current_figure('plot_bias_close_feature_{label}_{unique_id}.pdf')
return ax
def plot_compare_bic_collapsed_mixture_model(dataset, dataio=None):
'''
Check what models fits the data best using the BIC
As we have one BIC per subject (collapsed) and one BIC per subject / n_item (non-collapsed), we need to sum them properly somehow.
'''
collapsed_bic = np.empty((dataset['data_subject_split']['subjects_space'].size))
for subject, subject_data in dataset['collapsed_em_fits_subjects'].iteritems():
collapsed_bic[subject-1] = subject_data['bic']
separate_bic = np.empty((dataset['data_subject_split']['subjects_space'].size, dataset['data_subject_split']['nitems_space'].size))
for subject_i, subject_data in enumerate(dataset['em_fits_subjects_nitems'].values()):
for nitems_i, nitems_data in enumerate(subject_data.values()):
separate_bic[subject_i, nitems_i] = nitems_data['bic']
print 'Collapsed summed BIC: ', np.sum(collapsed_bic)
print 'Original non-collapsed BIC: ', np.sum(separate_bic)
f, ax = plt.subplots()
ax.plot(np.sum(separate_bic, axis=-1), collapsed_bic, 'o', markersize=10)
ixx = np.linspace(collapsed_bic.min()*0.95, collapsed_bic.max()*1.05, 100)
ax.plot(ixx, ixx, '--k')
ax.set_aspect('equal')
ax.set_xlabel('Non-collapsed BIC per subject')
ax.set_ylabel('Collapsed BIC per subject')
f.canvas.draw()
return separate_bic, collapsed_bic
def plot_collapsed_em_mixtures(dataset, dataio=None, use_sem=True):
'''
Do plots for the mixture models and kappa
'''
T_space_exp = dataset['data_subject_split']['nitems_space']
f, ax = plt.subplots()
if use_sem:
errorbars = 'sem'
else:
errorbars = 'std'
# Mixture probabilities
utils.plot_mean_std_area(T_space_exp, dataset['collapsed_em_fits']['mean']['mixt_target'], np.ma.masked_invalid(dataset['collapsed_em_fits'][errorbars]['mixt_target']).filled(0.0), xlabel='Number of items', ylabel="Mixture probabilities", ax_handle=ax, linewidth=3, fmt='o-', markersize=5, label='Target')
utils.plot_mean_std_area(T_space_exp, np.ma.masked_invalid(dataset['collapsed_em_fits']['mean']['mixt_nontargets']).filled(0.0), np.ma.masked_invalid(dataset['collapsed_em_fits'][errorbars]['mixt_nontargets']).filled(0.0), xlabel='Number of items', ylabel="Mixture probabilities", ax_handle=ax, linewidth=3, fmt='o-', markersize=5, label='Nontarget')
utils.plot_mean_std_area(T_space_exp, dataset['collapsed_em_fits']['mean']['mixt_random'], np.ma.masked_invalid(dataset['collapsed_em_fits'][errorbars]['mixt_random']).filled(0.0), xlabel='Number of items', ylabel="Mixture probabilities", ax_handle=ax, linewidth=3, fmt='o-', markersize=5, label='Random')
ax.legend(prop={'size':15})
ax.set_title('Mixture model for Collapsed EM fit %s' % dataset['name'])
ax.set_xlim([1.0, T_space_exp.max()])
ax.set_ylim([0.0, 1.1])
ax.set_xticks(range(1, T_space_exp.max()+1))
ax.set_xticklabels(range(1, T_space_exp.max()+1))
f.canvas.draw()
if dataio is not None:
dataio.save_current_figure('collapsedemfits_mixtures_{label}_{unique_id}.pdf')
# Kappa
f, ax = plt.subplots()
ax = utils.plot_mean_std_area(T_space_exp, dataset['collapsed_em_fits']['mean']['kappa'], np.ma.masked_invalid(dataset['collapsed_em_fits'][errorbars]['kappa']).filled(0.0), linewidth=3, fmt='o-', markersize=8, ylabel='Experimental data', ax_handle=ax, label='Kappa')
ax.legend(prop={'size':15})
ax.set_title('Kappa for Collapsed EM fit %s' % dataset['name'])
ax.set_xlim([0.9, T_space_exp.max()+0.1])
ax.set_ylim([0.0, np.max(dataset['collapsed_em_fits']['mean']['kappa'])*1.1])
ax.set_xticks(range(1, T_space_exp.max()+1))
ax.set_xticklabels(range(1, T_space_exp.max()+1))
ax.get_figure().canvas.draw()
if dataio is not None:
dataio.save_current_figure('collapsedemfits_kappa_{label}_{unique_id}.pdf')
def plot_compare_bic_collapsed_mixture_model_sequential(dataset, dataio=None):
'''
SEQUENTIAL GORGO 11 VERSION
Check what models fits the data best using the BIC
As we have one BIC per subject (collapsed) and one BIC per subject / n_item (non-collapsed), we need to sum them properly somehow.
'''
bic_collapsed_subjects_trecall = np.array([dataset['collapsed_em_fits_trecall']['values'][trecall]['bic'] for trecall in dataset['data_subject_split']['nitems_space']]).T
bic_collapsed_subjects_nitems = np.array([dataset['collapsed_em_fits_nitems']['values'][nitems]['bic'] for nitems in dataset['data_subject_split']['nitems_space']]).T
# em_fits_subjects_nitems_trecall
# em_fits_nitems_trecall
# em_fits_subjects_nitems
bic_separate_subjects_nitems_trecall = np.nan*np.empty((dataset['subject_size'], dataset['n_items_size'], dataset['n_items_size']))
ll_separate_subjects_nitems_trecall = np.nan*np.empty((dataset['subject_size'], dataset['n_items_size'], dataset['n_items_size']))
for subject_i, subject in enumerate(dataset['data_subject_split']['subjects_space']):
for n_items_i, n_items in enumerate(dataset['data_subject_split']['nitems_space']):
for t_i, trecall in enumerate(np.arange(1, n_items + 1)):
trecall_i = n_items - trecall
bic_separate_subjects_nitems_trecall[subject_i, n_items_i, t_i] = dataset['em_fits_subjects_nitems_trecall'][subject_i, n_items_i, trecall_i]['bic']
ll_separate_subjects_nitems_trecall[subject_i, n_items_i, t_i] = dataset['em_fits_subjects_nitems_trecall'][subject_i, n_items_i, trecall_i]['train_LL']
# Full double powerlaw model
bic_collapsed_subjects_doublepowerlaw = np.array([dataset['collapsed_em_fits_doublepowerlaw_subjects'][subj]['bic'] for subj in dataset['data_subject_split']['subjects_space']]).T
ll_collapsed_subjects_doublepowerlaw = np.array([dataset['collapsed_em_fits_doublepowerlaw_subjects'][subj]['train_LL'] for subj in dataset['data_subject_split']['subjects_space']]).T
print 'Collapsed trecall summed BIC: ', np.sum(bic_collapsed_subjects_trecall)
print 'Collapsed nitems summed BIC: ', np.sum(bic_collapsed_subjects_nitems)
print '\n\nCollapsed double powerlaw BIC: ', np.sum(bic_collapsed_subjects_doublepowerlaw)
print 'Original non-collapsed BIC: ', np.nansum(bic_separate_subjects_nitems_trecall)
print '\nCollapsed double powerlaw LL: ', np.sum(ll_collapsed_subjects_doublepowerlaw)
print 'Original non-collapsed LL: ', np.nansum(ll_separate_subjects_nitems_trecall)
bic_separate_subjects_trecall = np.nansum(bic_separate_subjects_nitems_trecall, axis=-2)
bic_separate_subjects_nitems = np.nansum(bic_separate_subjects_nitems_trecall, axis=-1)
# Plot Collapsed trecall vs separate (all)
if False:
f, axes = plt.subplots(nrows=1, ncols=2)
axes[0].plot(bic_separate_subjects_trecall.flatten(), bic_collapsed_subjects_trecall.flatten(), 'o', markersize=10)
ixx = np.linspace(min(bic_collapsed_subjects_trecall.min(), bic_separate_subjects_trecall.min())*0.95, max(bic_collapsed_subjects_trecall.max(), bic_separate_subjects_trecall.max())*1.05, 100)
axes[0].plot(ixx, ixx, '--k')
axes[0].set_aspect('equal')
axes[0].set_xlabel('Non-collapsed BIC, subject/nitem')
axes[0].set_ylabel('Collapsed trecall BIC, subject/nitem')
f.canvas.draw()
# Plot Collapsed nitems vs separate (all)
axes[1].plot(bic_separate_subjects_nitems.flatten(), bic_collapsed_subjects_nitems.flatten(), 'o', markersize=10)
ixx = np.linspace(min(bic_collapsed_subjects_nitems.min(), bic_separate_subjects_nitems.min())*0.95, max(bic_collapsed_subjects_nitems.max(), bic_separate_subjects_nitems.max())*1.05, 100)
axes[1].plot(ixx, ixx, '--k')
axes[1].set_aspect('equal')
axes[1].set_xlabel('Non-collapsed BIC, subject/nitem')
axes[1].set_ylabel('Collapsed nitems BIC, subject/nitem')
f.canvas.draw()
if dataio is not None:
dataio.save_current_figure('bic_separate_vs_collapsed_subjnitems_{label}_{unique_id}.pdf')
# Now do one summed BIC per subject
if False:
bic_collapsed_subjects_trecall_sum = np.nansum(bic_collapsed_subjects_trecall, axis=-1)
bic_collapsed_subjects_nitems_sum = np.nansum(bic_collapsed_subjects_nitems, axis=-1)
bic_separate_subjects_trecall_sum = np.nansum(bic_separate_subjects_trecall, axis=-1)
bic_separate_subjects_nitems_sum = np.nansum(bic_separate_subjects_nitems, axis=-1)
# Plot Collapsed trecall vs separate (per subject)
f, axes = plt.subplots(nrows=2, ncols=2, figsize=(15, 15))
axes[0, 0].plot(bic_separate_subjects_trecall_sum.flatten(), bic_collapsed_subjects_trecall_sum.flatten(), 'o', markersize=10)
ixx = np.linspace(min(bic_collapsed_subjects_trecall_sum.min(), bic_separate_subjects_trecall_sum.min())*0.95, max(bic_collapsed_subjects_trecall_sum.max(), bic_separate_subjects_trecall_sum.max())*1.05, 100)
axes[0, 0].plot(ixx, ixx, '--k')
axes[0, 0].set_aspect('equal')
axes[0, 0].set_xlabel('Non-collapsed BIC')
axes[0, 0].set_ylabel('Collapsed trecall BIC')
f.canvas.draw()
# Plot Collapsed nitems vs separate (per subject)
axes[0, 1].plot(bic_separate_subjects_nitems_sum.flatten(), bic_collapsed_subjects_nitems_sum.flatten(), 'o', markersize=10)
ixx = np.linspace(min(bic_collapsed_subjects_nitems_sum.min(), bic_separate_subjects_nitems_sum.min())*0.95, max(bic_collapsed_subjects_nitems_sum.max(), bic_separate_subjects_nitems_sum.max())*1.05, 100)
axes[0, 1].plot(ixx, ixx, '--k')
axes[0, 1].set_aspect('equal')
axes[0, 1].set_xlabel('Non-collapsed BIC')
axes[0, 1].set_ylabel('Collapsed nitems BIC')
f.canvas.draw()
if dataio is not None:
dataio.save_current_figure('bic_separate_vs_collapsed_subjects_{label}_{unique_id}.pdf')
# Plot double powerlaw vs collapsed nitems
axes[1, 0].plot(bic_collapsed_subjects_nitems_sum.flatten(), bic_collapsed_subjects_doublepowerlaw, 'o', markersize=10)
ixx = np.linspace(min(bic_collapsed_subjects_nitems_sum.min(), bic_collapsed_subjects_doublepowerlaw.min())*0.95, max(bic_collapsed_subjects_nitems_sum.max(), bic_collapsed_subjects_doublepowerlaw.max())*1.05, 100)
axes[1, 0].plot(ixx, ixx, '--k')
axes[1, 0].set_aspect('equal')
axes[1, 0].set_ylabel('Collapsed double powerlaw BIC')
axes[1, 0].set_xlabel('Collapsed nitems BIC')
f.canvas.draw()
# Plot double powerlaw vs collapsed trecall
axes[1, 1].plot(bic_collapsed_subjects_trecall_sum.flatten(), bic_collapsed_subjects_doublepowerlaw, 'o', markersize=10)
ixx = np.linspace(min(bic_collapsed_subjects_trecall_sum.min(), bic_collapsed_subjects_doublepowerlaw.min())*0.95, max(bic_collapsed_subjects_trecall_sum.max(), bic_collapsed_subjects_doublepowerlaw.max())*1.05, 100)
axes[1, 1].plot(ixx, ixx, '--k')
axes[1, 1].set_aspect('equal')
axes[1, 1].set_ylabel('Collapsed double powerlaw BIC')
axes[1, 1].set_xlabel('Collapsed trecall BIC')
f.canvas.draw()
# Compare double-powerlaw vs original, per subject and nitems
if True:
bic_separate_subjects = np.nansum(
np.nansum(bic_separate_subjects_nitems_trecall, axis=-1),
axis=-1)
f, ax = plt.subplots()
ax.plot(
bic_separate_subjects.flatten(),
bic_collapsed_subjects_doublepowerlaw,
'o', markersize=10)
ixx = np.linspace(
min(bic_separate_subjects.min(),
bic_collapsed_subjects_doublepowerlaw.min())*0.95,
max(bic_separate_subjects.max(),
bic_collapsed_subjects_doublepowerlaw.max())*1.05,
100)
ax.plot(ixx, ixx, '--k')
ax.set_aspect('equal')
ax.set_xlabel('Original model BIC')
ax.set_ylabel('Collapsed double power-law BIC')
f.canvas.draw()
if dataio is not None:
dataio.save_current_figure('bic_separate_vs_collapsed_doublepowerlaw_{label}_{unique_id}.pdf')
# Works well. Overall quite a huge improvement in BIC, as I thought.
def plot_sequential_histograms_errors(dataset, dataio=None):
'''
Create subplots showing histograms of errors to targets and nontargets
'''
angle_space = np.linspace(-np.pi, np.pi, 51)
#### Histograms of errors / errors to nontargets, collapsing across subjects
if True:
f1, axes_target = plt.subplots(
ncols=dataset['n_items_size'],
nrows=dataset['n_items_size'],
figsize=(5, 5))
f2, axes_nontarget = plt.subplots(
ncols=dataset['n_items_size'],
nrows=dataset['n_items_size'],
figsize=(5, 5))
for n_items_i, n_items in enumerate(np.unique(dataset['n_items'])):
for trecall_i, trecall in enumerate(np.unique(dataset['n_items'])):
if trecall <= n_items:
utils.hist_angular_data(
dataset['errors_nitems_trecall'][n_items_i, trecall_i],
bins=angle_space,
# title='N=%d, trecall=%d' % (n_items, trecall),
norm='density',
ax_handle=axes_target[n_items_i, trecall_i],
pretty_xticks=False)
axes_target[n_items_i, trecall_i].set_ylim([0., 1.4])
axes_target[n_items_i, trecall_i].xaxis.set_major_locator(
plt.NullLocator())
axes_target[n_items_i, trecall_i].yaxis.set_major_locator(
plt.NullLocator())
if n_items > 1:
utils.hist_angular_data(
utils.dropnan(dataset['errors_nontarget_nitems_trecall'][n_items_i, trecall_i]),
bins=angle_space,
# title='Nontarget %s N=%d' % (dataset['name'], n_items),
norm='density',
ax_handle=axes_nontarget[n_items_i, trecall_i],
pretty_xticks=False)
axes_nontarget[n_items_i, trecall_i].set_ylim([0., 0.3])
axes_nontarget[
n_items_i, trecall_i].xaxis.set_major_locator(
plt.NullLocator())