-
Notifications
You must be signed in to change notification settings - Fork 1
/
norms.R
2203 lines (1658 loc) · 90.8 KB
/
norms.R
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
#
# Code for a psycholinguistic study about conceptual modality, part of P. Bernabeu's
# MPhil thesis. 'Modality exclusivity norms for 747 properties and concepts in Dutch:
# a replication of English'. More info at: https://goo.gl/Je4JGO.
# Contact: [email protected]
# READ-ME
# Ctrl+f may be used to search for specific parts of this code (for instance, PCA).
# The 'all.csv' file, created outside of R, in Excel, compiles all individual ratings.
# Dutch and English data are described in separate columns. All analyses separate for
# properties and concepts, except for a translation check.
# Stat tests (specifying treatment of English and Dutch norms): reliability analysis
# (only Dutch norms), Pearson?s correlation (norms independent and paired), one-sample
# t-test (norms independent), Principal Components Analysis (norms independent), ANOVA
# (norms paired), and multiple regression (norms independent).
# The code is extensively annotated, but some clarifications are in order. Subsetting is
# done throughout the code, and is essential due to the different norms (see 'normed'
# column: English, Dutch, or both). Subsetting is often done on the basis of variables
# that are unique to either norms, especially, 'Exclusivity' and 'English_Exclusivity_Lynott_Connell_2009_2013'.
# At first, the code must be run right from the top, as different objects bear the
# same name. In its entirety, it takes ~20 mins. Note the annotations for theoretical
# matters. Long variables are never presented entirely, but rather in sections or via
# summaries. Yet, the reader is invited to edit and present them entirely.
# Written on R version 3.2.2 (2015-08-14). This script markdown presents each code chunk
# followed by the results.
# INDEX
# Libraries: please ensure every library loads, and otherwise install it via
# install.packages("")
# Preprocessing
# Modality
# Sound-symbolism
setwd('C:/Users/Pablo/Dropbox/STUDIES/R/Experiment Data/Modality norms')
install.packages("gdata")
install.packages("GPArotation")
install.packages("psych")
install.packages("ggplot2")
install.packages("car")
install.packages("Rmisc")
install.packages("corpcor")
install.packages('contrast')
install.packages('doBy')
install.packages('ltm')
install.packages('MASS')
install.packages('QuantPsyc')
install.packages('qpcR')
install.packages('corpcor')
install.packages('lattice')
install.packages('car')
install.packages('pastecs')
install.packages('scales')
install.packages('reshape')
install.packages('arules')
install.packages('plyr')
install.packages('RColorBrewer')
install.packages('dplyr')
install.packages('gdata')
install.packages('gtools')
install.packages('Hmisc')
install.packages('png')
install.packages('ggrepel')
install.packages('irr')
install.packages('tibble')
library(ltm)
library(lattice)
library(psych)
library(car)
library(doBy)
library(contrast)
library(pastecs)
library(scales)
library(ggplot2)
library(reshape)
library(arules)
library(plyr)
library(RColorBrewer)
library(Rmisc)
library(corpcor)
library(GPArotation)
library(gdata)
library(QuantPsyc)
library(MASS)
library(qpcR)
library(dplyr)
library(gtools)
library(Hmisc)
library(png)
library(ggrepel)
library(irr)
library(tibble)
# Calculate average percentange of unresponded items, i.e., unknown. Since there are
# three ratings per word, and indeed the three were left blank whereever participants
# ignored some word, the calculation includes a division by 3 (besides overall mean,
# see specific percentage per file).
file1 = read.csv('file1_gral.csv')
file2 = read.csv('file2_gral.csv')
file3 = read.csv('file3_gral.csv')
file4 = read.csv('file4_gral.csv')
file5 = read.csv('file5_gral.csv')
file6 = read.csv('file6_gral.csv')
# What percentage of modality ratings was not provided by respondents? Missing divided by total:
( sum(is.na(file1)) + sum(is.na(file2)) + sum(is.na(file3)) +
sum(is.na(file4[,-c(1:4)])) + # first participant removed because they completed only first half survey
sum(is.na(file5)) + sum(is.na(file6)) ) /
( ncol(file1[,-1]) * nrow(file1[,-1]) +
ncol(file2[,-1]) * nrow(file2[,-1]) +
ncol(file3[,-1]) * nrow(file3[,-1]) +
ncol(file4[,-c(1:4)]) * nrow(file4[,-c(1:4)]) + # first participant removed as above
ncol(file5[,-1]) * nrow(file5[,-1]) +
ncol(file6[,-1]) * nrow(file6[,-1]) ) *
100 # percentage
# 3.82% missing ratings
#_________________________________________________________________________________
# Preprocessing:
# There were 9 files with different items (mostly unrepeated) for concepts and 10
# files for properties. They were completed in different proportions, with an
# average of eight participants per file.
# RELIABILITY ANALYSIS: In putting together the ratings from each respondent, this
# analysis allows to calculate the fit among those. In other words, is the mean
# realistic or forced? Two measures are provided. First, interitem consistency
# provides the fit among items independently of raters. Second, interrater
# reliability measures the fit among raters, independently of items. Interrater
# reliability is first calculated on the modality scores, and then at the level
# of dominant modalities.
# Concepts
all = read.csv('all.csv')
concs = all[all$cat == 'Concept',]
a_concs = concs[, c('a1', 'a2', 'a3', 'a4', 'a5', 'a6', 'a7', 'a8', 'a9', 'a10')]
h_concs = concs[, c('h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'h7', 'h8', 'h9', 'a10')]
v_concs = concs[, c('v1', 'v2', 'v3', 'v4', 'v5', 'v6', 'v7', 'v8', 'v9', 'a10')]
# Interitem consistency
psych::alpha(a_concs)
psych::alpha(h_concs)
psych::alpha(v_concs)
# a: .74
# h: .72
# v: .70
# Interrater reliability (Koo & Li, 2016)
a_concs = concs[, c('a1', 'a2', 'a3', 'a4', 'a5', 'a6', 'a7', 'a8', 'a9')]
h_concs = concs[, c('h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'h7', 'h8', 'h9')]
v_concs = concs[, c('v1', 'v2', 'v3', 'v4', 'v5', 'v6', 'v7', 'v8', 'v9')]
icc(a_concs, "oneway", "consistency")
icc(h_concs, "oneway", "consistency")
icc(v_concs, "oneway", "consistency")
# Now, look at the inter-rater agreement on the basis of the highest modality score for each word, namely
# the dominant modality. Wherever a tie occurs among several modalities, one of them is randomly selected.
all.Dutch = all[!is.na(all$word),]
rater = NA
id = NA # i.e., word ID
result = NA
results = data.frame(rater, id, result)
# Safer version of sample(), for use below
resample = function(x, ...) x[sample.int(length(x), ...)]
getMain =
function(rater, i.col){
for(i.row in 1:nrow(all.Dutch)){
# For cases in which no rating was provided to any modality, assign NA
if( length(which(all.Dutch[i.row, c(i.col)]==max(all.Dutch[i.row, c(i.col)]))) == 0) {
result = NA
# For tied modalities, randomly select one of them
} else if( !length(which(all.Dutch[i.row, c(i.col)]==max(all.Dutch[i.row, c(i.col)]))) > 1) {
result = ifelse( sample(which(all.Dutch[i.row, c(i.col)]==max(all.Dutch[i.row, c(i.col)])), size=1) == 1, 'Auditory',
ifelse( sample(which(all.Dutch[i.row, c(i.col)]==max(all.Dutch[i.row, c(i.col)])), size=1) == 2, 'Haptic',
'Visual' ) )
# For plain cases with one dominant modality, assign that modality
} else{
result = ifelse( which(all.Dutch[i.row, c(i.col)]==max(all.Dutch[i.row, c(i.col)])) == 1, 'Auditory',
ifelse( which(all.Dutch[i.row, c(i.col)]==max(all.Dutch[i.row, c(i.col)])) == 2, 'Haptic',
'Visual' ) )
}
results = rbind(results, c(rater, as.character(all.Dutch[i.row, 'id']), result))
}
results <= results[!is.na(results$rater),]
}
getMain(1, c('a1','h1','v1'))
getMain(2, c('a2','h2','v2'))
getMain(3, c('a3','h3','v3'))
getMain(4, c('a4','h4','v4'))
getMain(5, c('a5','h5','v5'))
getMain(6, c('a6','h6','v6'))
getMain(7, c('a7','h7','v7'))
getMain(8, c('a8','h8','v8'))
getMain(9, c('a9','h9','v9'))
getMain(10, c('a10','h10','v10'))
all = merge(all, results[results$rater==1, c('id', 'result')], by='id')
colnames(all)[colnames(all)=="result"] = 'main1'
all = merge(all, results[results$rater==2, c('id', 'result')], by='id')
colnames(all)[colnames(all)=="result"] = 'main2'
all = merge(all, results[results$rater==3, c('id', 'result')], by='id')
colnames(all)[colnames(all)=="result"] = 'main3'
all = merge(all, results[results$rater==4, c('id', 'result')], by='id')
colnames(all)[colnames(all)=="result"] = 'main4'
all = merge(all, results[results$rater==5, c('id', 'result')], by='id')
colnames(all)[colnames(all)=="result"] = 'main5'
all = merge(all, results[results$rater==6, c('id', 'result')], by='id')
colnames(all)[colnames(all)=="result"] = 'main6'
all = merge(all, results[results$rater==7, c('id', 'result')], by='id')
colnames(all)[colnames(all)=="result"] = 'main7'
all = merge(all, results[results$rater==8, c('id', 'result')], by='id')
colnames(all)[colnames(all)=="result"] = 'main8'
all = merge(all, results[results$rater==9, c('id', 'result')], by='id')
colnames(all)[colnames(all)=="result"] = 'main9'
all = merge(all, results[results$rater==10, c('id', 'result')], by='id')
colnames(all)[colnames(all)=="result"] = 'main10'
# Order columns
names(all)
all = all[,c(1:58, 67:76, 59:66)]
# Fleiss' Kappa (Fleiss, 1971)
kappam.fleiss(all[,c('main1', 'main2', 'main3', 'main4', 'main5', 'main6', 'main7',
'main8', 'main9', 'main10')], detail=TRUE)
# Properties
# Interitem consistency
props = all[all$cat == 'Property',]
a_props = props[, c('a1', 'a2', 'a3', 'a4', 'a5', 'a6', 'a7', 'a8', 'a9', 'a10')]
h_props = props[, c('h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'h7', 'h8', 'h9', 'h10')]
v_props = props[, c('v1', 'v2', 'v3', 'v4', 'v5', 'v6', 'v7', 'v8', 'v9', 'v10')]
psych::alpha(a_props)
psych::alpha(h_props)
psych::alpha(v_props)
# a: .78
# h: .70
# v: .85
# Interrater reliability (Koo & Li, 2016)
a_props = props[, c('a1', 'a2', 'a3', 'a4', 'a5', 'a6', 'a7', 'a8')]
h_props = props[, c('h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'h7', 'h8')]
v_props = props[, c('v1', 'v2', 'v3', 'v4', 'v5', 'v6', 'v7', 'v8')]
icc(a_props, "oneway", "consistency")
icc(h_props, "oneway", "consistency")
icc(v_props, "oneway", "consistency")
# _______________________________________________________________________________
# Read in the general norms file, 'all'
all = read.csv('all.csv')
# PROPERTIES
props = all[all$cat=='Property',]
nrow(props) # 336 Dutch + a few items from Lynott and Connell (2009) that weren't translated but were kept in to have all items from the three modalities in the PCA reanalysis.
# CONCEPTS
concs = all[all$cat=='Concept',]
nrow(concs) # 411 Dutch + a few items from Lynott and Connell (2013) that weren't translated but were kept in to have all items from the three modalities in the PCA reanalysis.
# Descriptives: perceptual strength
# Dutch
summaryBy(Perceptualstrength ~ cat, FUN=stat.desc, data=all)
# English
summaryBy(English_Perceptualstrength_Lynott_Connell_2009_2013 ~ cat, FUN=stat.desc,
data=all[!is.na(all$English_Perceptualstrength_Lynott_Connell_2009_2013),])
# The differences make sense considering the sampling method applied in the different
# norms. The difference between the English and the Dutch norms is a bit larger for
# the concept norms because these norms were sampled differently. While the concepts
# tested in the English norms were compiled regardless of their potential modalities,
# the Dutch norms were entirely compiled with a view to potential modality.
# Correlations
# PROPERTIES
# Modalities
rcor.test(props[, c('Auditory', 'English_Auditory_Lynott_Connell_2009_2013')], use = 'complete.obs')
rcor.test(props[, c('Haptic', 'English_Haptic_Lynott_Connell_2009_2013')], use = 'complete.obs')
rcor.test(props[, c('Visual', 'English_Visual_Lynott_Connell_2009_2013')], use = 'complete.obs')
# Significant, large correlations ranging from .69 to .80
# Exclusivity
rcor.test(props[, c('Exclusivity', 'English_Exclusivity_Lynott_Connell_2009_2013')], use = 'complete.obs')
# Medium-sized corr Eng-Dutch
# CONCEPTS
# Modalities
rcor.test(concs[, c('Auditory', 'English_Auditory_Lynott_Connell_2009_2013')], use = 'complete.obs')
rcor.test(concs[, c('Haptic', 'English_Haptic_Lynott_Connell_2009_2013')], use = 'complete.obs')
rcor.test(concs[, c('Visual', 'English_Visual_Lynott_Connell_2009_2013')], use = 'complete.obs')
# Significant, large correlations ranging from .63 to .69
# Exclusivity
rcor.test(concs[, c('Exclusivity', 'English_Exclusivity_Lynott_Connell_2009_2013')], use = 'complete.obs')
# Medium-sized corr Eng-Dutch
# Descriptives: M, SD, SE...
# English
psych::describe(props$English_Auditory_Lynott_Connell_2009_2013)
psych::describe(props$English_Haptic_Lynott_Connell_2009_2013)
psych::describe(props$English_Visual_Lynott_Connell_2009_2013)
psych::describe(concs$English_Auditory_Lynott_Connell_2009_2013)
psych::describe(concs$English_Haptic_Lynott_Connell_2009_2013)
psych::describe(concs$English_Visual_Lynott_Connell_2009_2013)
stat.desc(props$English_Auditory_Lynott_Connell_2009_2013)
stat.desc(props$English_Haptic_Lynott_Connell_2009_2013)
stat.desc(props$English_Visual_Lynott_Connell_2009_2013)
stat.desc(concs$English_Auditory_Lynott_Connell_2009_2013)
stat.desc(concs$English_Haptic_Lynott_Connell_2009_2013)
stat.desc(concs$English_Visual_Lynott_Connell_2009_2013)
# Dutch
psych::describe(props$Auditory)
psych::describe(props$Haptic)
psych::describe(props$Visual)
psych::describe(concs$Auditory)
psych::describe(concs$Haptic)
psych::describe(concs$Visual)
stat.desc(props$Auditory)
stat.desc(props$Haptic)
stat.desc(props$Visual)
stat.desc(concs$Auditory)
stat.desc(concs$Haptic)
stat.desc(concs$Visual)
# Sample sizes for English and Dutch
nrow(props[!is.na(props$English_Exclusivity_Lynott_Connell_2009_2013),]) # total properties w/ English norms (Lynott & Connell, 2013) = 343
nrow(concs[!is.na(concs$English_Exclusivity_Lynott_Connell_2009_2013),]) # total concepts w/ English norms (Lynott & Connell, 2013) = 392
nrow(props[!is.na(props$Exclusivity),]) # total properties w/ Dutch norms = 336
nrow(props[props$main=='Auditory' & !is.na(props$Exclusivity),])
nrow(props[props$main=='Haptic' & !is.na(props$Exclusivity),])
nrow(props[props$main=='Visual' & !is.na(props$Exclusivity),])
nrow(props[!is.na(concs$Exclusivity),]) # total concepts w/ Dutch norms = 411
nrow(props[concs$main=='Auditory' & !is.na(concs$Exclusivity),])
nrow(props[concs$main=='Haptic' & !is.na(concs$Exclusivity),])
nrow(props[concs$main=='Visual' & !is.na(concs$Exclusivity),])
# _______________________________________________________________
# Relation between modality strength, dominant modalities, and mod exclusivitY
# ENGLISH
# properties
summaryBy(English_Auditory_Lynott_Connell_2009_2013 ~ English_Main_Lynott_Connell_2009_2013, data=props, FUN=mean)
summaryBy(English_Haptic_Lynott_Connell_2009_2013 ~ English_Main_Lynott_Connell_2009_2013, data=props, FUN=mean)
summaryBy(English_Visual_Lynott_Connell_2009_2013 ~ English_Main_Lynott_Connell_2009_2013, data=props, FUN=mean)
summaryBy(English_Exclusivity_Lynott_Connell_2009_2013 ~ English_Main_Lynott_Connell_2009_2013, data=props, FUN=mean)
# concepts
summaryBy(English_Auditory_Lynott_Connell_2009_2013 ~ English_Main_Lynott_Connell_2009_2013, data=concs, FUN=mean)
summaryBy(English_Haptic_Lynott_Connell_2009_2013 ~ English_Main_Lynott_Connell_2009_2013, data=concs, FUN=mean)
summaryBy(English_Visual_Lynott_Connell_2009_2013 ~ English_Main_Lynott_Connell_2009_2013, data=concs, FUN=mean)
summaryBy(English_Exclusivity_Lynott_Connell_2009_2013 ~ English_Main_Lynott_Connell_2009_2013, data=concs, FUN=mean)
# DUTCH
# properties
summaryBy(Auditory ~ main, data=props, FUN=mean)
summaryBy(Haptic ~ main, data=props, FUN=mean)
summaryBy(Visual ~ main, data=props, FUN=mean)
summaryBy(Exclusivity ~ main, data=props, FUN=mean)
# concepts
summaryBy(Auditory ~ main, data=concs, FUN=mean)
summaryBy(Haptic ~ main, data=concs, FUN=mean)
summaryBy(Visual ~ main, data=concs, FUN=mean)
summaryBy(Exclusivity ~ main, data=concs, FUN=mean)
# RESULTS: both languages strongly related on individual modalities and on
# exclusivity. Correlations among modalities replicate previous norms, with visual
# and haptic items related, and auditory ones independent.
# Yet, there is clearly a greater exclusivity in the English norms.
# Properties
psych::describe(props$English_Exclusivity_Lynott_Connell_2009_2013) # M = 0.48
psych::describe(props$Exclusivity) # M = 0.40
# Concepts
psych::describe(concs$English_Exclusivity_Lynott_Connell_2009_2013) # M = 0.40
psych::describe(concs$Exclusivity) # M = 0.29
# Indeed lower exclusivity and higher SD for Dutch items >> Check significance
# Because the English and the Dutch norms are paired, the difference has to be
# checked through a one-sample t-test, checking the mean of one language against
# the other language (see further below).
# Correlations among modalities within each category and language:
# ENGLISH
# PROPERTIES
rcor.test(props[, c('English_Auditory_Lynott_Connell_2009_2013', 'English_Haptic_Lynott_Connell_2009_2013', 'English_Visual_Lynott_Connell_2009_2013', 'English_Exclusivity_Lynott_Connell_2009_2013')], use =
'complete.obs')
corr3 = rcor.test(props[, c('English_Auditory_Lynott_Connell_2009_2013', 'English_Haptic_Lynott_Connell_2009_2013', 'English_Visual_Lynott_Connell_2009_2013', 'English_Exclusivity_Lynott_Connell_2009_2013')],
use = 'complete.obs')
#write.csv(corr3$cor.mat, file = "corr3.csv",na="") # saved for manuscript
# CONCEPTS
rcor.test(concs[, c('English_Auditory_Lynott_Connell_2009_2013', 'English_Haptic_Lynott_Connell_2009_2013', 'English_Visual_Lynott_Connell_2009_2013', 'English_Exclusivity_Lynott_Connell_2009_2013')], use = 'complete.obs')
corr4 = rcor.test(concs[, c('English_Auditory_Lynott_Connell_2009_2013', 'English_Haptic_Lynott_Connell_2009_2013', 'English_Visual_Lynott_Connell_2009_2013', 'English_Exclusivity_Lynott_Connell_2009_2013')], use =
'complete.obs')
#write.csv(corr4$cor.mat, file = "corr4.csv",na="") # saved for manuscript
# DUTCH
# PROPERTIES
rcor.test(props[, c('Auditory', 'Haptic', 'Visual', 'Exclusivity')], use =
'complete.obs')
corr1 = rcor.test(props[, c('Auditory', 'Haptic', 'Visual', 'Exclusivity')],
use = 'complete.obs')
#write.csv(corr1$cor.mat, file = "corr1.csv",na="") # saved for manuscript
# CONCEPTS
rcor.test(concs[, c('Auditory', 'Haptic', 'Visual', 'Exclusivity')], use =
'complete.obs')
corr2 = rcor.test(concs[, c('Auditory', 'Haptic', 'Visual', 'Exclusivity')], use =
'complete.obs')
#write.csv(corr2$cor.mat, file = "corr2.csv",na="") # saved for manuscript
# Statistical tests for those differences
# Yet the same again, but now with a statistical significance test
# ENGLISH
# Setting contrasts based on means
contrasts(all$English_Main_Lynott_Connell_2009_2013) = cbind(c(2,0,-2), c(-1,2,-1))
# (1) Aud vs Vis; (2) Hap vs Aud-&-Vis
contrasts(all$English_Main_Lynott_Connell_2009_2013)
fitt = aov(English_Exclusivity_Lynott_Connell_2009_2013 ~ English_Main_Lynott_Connell_2009_2013 * cat, data=all)
plot(fitt)
summary(fitt)
drop1(fitt,~.,test="F")
Anova(fitt)
Anova(fitt, type = "II")
Anova(fitt, type = "III")
summary.lm(fitt)
# RESULTS: English properties with more exclusivity than concepts(***)
# Contrasts: (1) Aud vs Vis (*)
# (2) Haptic words show less exclusivity than auditory and visual ones within
# properties but not within concepts (*)
# DUTCH
# Setting contrasts based on means
contrasts(all$main) = cbind(c(2,0,-2), c(-1,2,-1))
# (1) Aud vs Vis; (2) Hap vs Aud-&-Vis
contrasts(all$main)
fitt = aov(Exclusivity ~ main * cat, data=all)
plot(fitt) # must click over the plot several times in order to continue
summary(fitt)
drop1(fitt,~.,test="F")
Anova(fitt)
Anova(fitt, type = "II")
Anova(fitt, type = "III")
summary.lm(fitt)
# RESULTS: Dutch properties with more exclusivity than concepts(***)
# Contrasts: (1) Aud vs Vis (non-sig)
# (2) Haptic words show less exclusivity than auditory and visual ones within
# properties but not within concepts (**)
# Overall, these results stem from the nature of human perception. What exclusivity
# seems to be indexing is the degree to which percepts will naturally co-occur. Thus,
# visual and auditory words have relatively higher exclusivities because what we
# see and hear often stands on its own. We can often see thing but not hear or touch
# them. By the same token, we often hear things that we cannot see or touch. Now, in
# contrast, if we can touch something, we likely can see and hear it too--hence the
# low exclusivity of haptic items.
# SAME PLOT-WISE:
# Barplot of exclusivity percentiles within modalities for Dutch items (as in
# Van Dantzig et al., 2011, but separately for properties and concepts)
all=read.csv('all.csv')
allNL = all[!is.na(all$main),]
allNL$Range = floor(allNL$Exclusivity * 4)
allNL$Range = mapvalues(allNL$Range, from = c(0, 1, 2, 3, 4),
to = c("0-20%", "20-40%", "40-60%", "60-80%", "80-100%"))
allNL$cat = recode(allNL$cat, Concept = "Concepts", Property = "Properties")
# Set order to display properties first, instead of alphabetical 'Concepts-Properties' order
allNL$cat = factor(allNL$cat, levels=rev(levels(allNL$cat)))
savedplot = ggplot(allNL) +
geom_bar(mapping = aes(x = main, fill = Range), position = position_stack(reverse = TRUE)) +
scale_fill_grey(start=.9, end=0, labels = c("0-20%", "20-40%", "40-60%", "60-80%", "80-100%"),
guide = guide_legend(reverse = TRUE, override.aes = list(size = 11))) +
scale_x_discrete(expand = c(.24,0)) + scale_y_continuous(expand = expand_scale(mult = c(0, .05))) +
facet_grid(. ~ cat) + labs(fill = "Modality\nExclusivity", x = 'Dominant Modality', y = 'Number of Words') +
theme_bw() + theme(legend.position = c(.17, .61), legend.title = element_text(size = 20, face = 'bold'),
legend.text = element_text(size = 18), legend.background = element_rect(fill=alpha('white', 0)),
axis.title = element_text(size = 21, face = "bold"), axis.text = element_text(size = 19),
panel.border = element_blank(), panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), axis.line = element_line(colour = "black"),
strip.text = element_text(size = 21, face = 'bold', vjust = .7,
margin = margin(.3, 0, .3, 0, "cm")),
axis.ticks.x = element_blank())
savedplot
# Export to disk
png(file="stacked_exc.png", units="in", width=10, height=9, res=1000)
savedplot
dev.off()
# Same plot for the English items of Lynott and Connell (witout gustatory and olfactory items)
allENG = allENG[!is.na(all$English_Main_Lynott_Connell_2009_2013),]
allENG$Range = floor(allENG$Exclusivity * 4)
allENG$Range = mapvalues(allENG$Range, from = c(0, 1, 2, 3, 4),
to = c("0-20%", "20-40%", "40-60%", "60-80%", "80-100%"))
allENG = allENG[!is.na(allENG$Range),]
allENG$cat = recode(allENG$cat, Concept = "Concepts", Property = "Properties")
savedplot = ggplot(allENG) +
geom_bar(mapping = aes(x = main, fill = Range), position = position_stack(reverse = TRUE)) +
scale_fill_grey(start=.9, end=0, labels = c("0-20%", "20-40%", "40-60%", "60-80%", "80-100%"),
guide = guide_legend(reverse = TRUE, override.aes = list(size = 11))) +
scale_x_discrete(expand = c(.24,0)) + scale_y_continuous(expand = expand_scale(mult = c(0, .05))) +
facet_grid(. ~ cat) + labs(fill = "Modality\nExclusivity", x = 'Dominant Modality', y = 'Number of Words') +
theme_bw() + theme(legend.position = c(.17, .57), legend.title = element_text(size = 20, face = 'bold'),
legend.text = element_text(size = 18), legend.background = element_rect(fill=alpha('white', 0)),
axis.title = element_text(size = 21, face = "bold"), axis.text = element_text(size = 19),
panel.border = element_blank(), panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), axis.line = element_line(colour = "black"),
strip.text = element_text(size = 21, face = 'bold', vjust = .7,
margin = margin(.3, 0, .3, 0, "cm")),
axis.ticks.x = element_blank())
savedplot
# Export to disk
png(file="stacked_exc.png", units="in", width=6, height=6, res=1000)
savedplot
dev.off()
# Comparison English Dutch on exclusivity
# Properties
t.test(props$English_Exclusivity_Lynott_Connell_2009_2013, mu = 0.40)
# The difference is considerable, t(734) = 18.8, p < .001
# dz = t/vn = 0.47
# Concepts
t.test(concs$English_Exclusivity_Lynott_Connell_2009_2013, mu = 0.29)
# The difference is considerable, t(734) = 18.8, p < .001
# dz = t/vn = 0.83
# ___________________________________________________________________________
# RELATION AMONG MODALITIES
# Below is a Principal Components Analysis (PCA) with plots. Firstly it is performed
# on the Dutch norms, and then on Lynott and Connell's (2009, 2013) English norms
# (leaving out gustatory and olfactory scores and words).
all = read.csv('all.csv')
nrow(all) # 747 used in Dutch norms + 12 from English not used
# ON ENGLISH NORMS
# PCA plotting on the English norms, as based on Lynott and Connell's supplementary materials
# (http://www.lancaster.ac.uk/people/connelll/lab/norms.html).
# English Properties
# check conditions for a PCA
# matrix
eng_prop = all[all$cat == 'Property', c('English_Auditory_Lynott_Connell_2009_2013', 'English_Haptic_Lynott_Connell_2009_2013', 'English_Visual_Lynott_Connell_2009_2013')]
nrow(eng_prop)
eng_prop_matrix = cor(eng_prop, use = 'complete.obs')
eng_prop_matrix
round(eng_prop_matrix, 2)
# OK: correlations good for a PCA, with enough < .3
# now on the raw vars:
nrow(eng_prop)
cortest.bartlett(eng_prop)
# GOOD: Bartlett's test significant
# KMO: Kaiser-Meyer-Olkin Measure of Sampling Adequacy
KMO(eng_prop_matrix)
# Result: .56 = mediocre. PCA not strongly recommended. But we still do it
# because the purpose is graphical only.
# check determinant
det(eng_prop_matrix)
# GOOD: > 0.00001
# start off with unrotated PCA
pc1_eng_prop = psych::principal(eng_prop, nfactors = 3, rotate = "none")
pc1_eng_prop
# RESULT: Extract either one PC, acc to Kaiser's criterion, or two RCs, acc to
# Joliffe's (Field, Miles, & Field, 2012)
# Unrotated: scree plot
plot(pc1_eng_prop$values, type = "b")
# Result: again one or two RCs should be extracted
# Now with varimax rotation, Kaiser-normalized (by default)
pc2_eng_prop = psych::principal(eng_prop, nfactors = 2, rotate = "varimax", scores = TRUE)
pc2_eng_prop
pc2_eng_prop$loadings
# two components are good, as they both have eigenvalues over 1
pc2_eng_prop$residual
pc2_eng_prop$fit
pc2_eng_prop$communality
# Results based on a Kaiser-normalizalized orthogonal (varimax) rotation
# (by default in psych::stats). Residuals bad: more than 50% have absolute
# values > 0.05. Model fit good, > .90. Communalities good, all > .7.
# subset and add PCs
eng_props = all[all$cat == 'Property', ]
nrow(eng_props)
eng_props = cbind(eng_props, pc2_eng_prop$scores)
nrow(eng_props)
head(eng_props)
# Finally, plot
# Set sample words to show on plot (first word in each modality)
auditory_w = as.character(sort(eng_props[eng_props$English_Main_Lynott_Connell_2009_2013=='Auditory', 'English_Word_Lynott_Connell_2009_2013'])[1])
haptic_w = as.character(sort(eng_props[eng_props$English_Main_Lynott_Connell_2009_2013=='Haptic', 'English_Word_Lynott_Connell_2009_2013'])[1])
visual_w = as.character(sort(eng_props[eng_props$English_Main_Lynott_Connell_2009_2013=='Visual', 'English_Word_Lynott_Connell_2009_2013'])[1])
w_set = c(auditory_w, haptic_w, visual_w)
eng_props$English_Main_Lynott_Connell_2009_2013 = recode(eng_props$English_Main_Lynott_Connell_2009_2013, Auditory = "a", Haptic = "h", Visual = "v")
Engprops = ggplot(eng_props,
aes(RC1, RC2, label = as.character(English_Main_Lynott_Connell_2009_2013))) + stat_density2d (color = "gray87") +
geom_text(size = ifelse(eng_props$English_Word_Lynott_Connell_2009_2013 %in% w_set, 12, 7),
fontface = ifelse(eng_props$English_Word_Lynott_Connell_2009_2013 %in% w_set, 'bold', 'plain')) +
geom_point(data=eng_props[eng_props$English_Word_Lynott_Connell_2009_2013 %in% w_set,], pch=21, fill=NA, size=14, stroke=2, alpha=.6) +
labs(x = "Varimax-rotated Principal Component 1", y = "Varimax-rotated Principal Component 2") + theme_bw() +
theme( plot.background = element_blank(), panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.border = element_blank(),
axis.line = element_line(color = 'black'),
axis.title.x = element_text(colour = 'black', size = 23, margin=margin(15,15,15,15)),
axis.title.y = element_text(colour = 'black', size = 23, margin=margin(15,15,15,15)),
axis.text.x = element_text(size=16), axis.text.y = element_text(size=16),
plot.title = element_text(hjust = 0.5, size = 32, face = "bold", margin=margin(15,15,15,15)),
plot.subtitle = element_text(hjust = 0.5, size = 20, margin=margin(2,15,15,15)) ) +
geom_label_repel(data = eng_props[eng_props$English_Word_Lynott_Connell_2009_2013 %in% w_set,], aes(label = English_Word_Lynott_Connell_2009_2013), size = 8,
alpha = 0.77, color = 'black', box.padding = 1.5 )
plot(Engprops) # ! THE PLOT IS SHOWN BADLY ON HERE. PLEASE SEE THE SAVED PLOTS
# Now to save, run first line below and return to keep running. See your folder.
#png(file="Engprops_highres.png", units="in", width=13, height=13, res=900)
#plot(Engprops)
#dev.off()
# Adjust for combined plots:
Engprops4 = ggplot(eng_props,
aes(RC1, RC2, label = as.character(English_Main_Lynott_Connell_2009_2013))) + stat_density2d (color = "gray87") +
geom_text(size = ifelse(eng_props$English_Word_Lynott_Connell_2009_2013 %in% w_set, 8, 5),
fontface = ifelse(eng_props$English_Word_Lynott_Connell_2009_2013 %in% w_set, 'bold', 'plain')) +
geom_point(data=eng_props[eng_props$English_Word_Lynott_Connell_2009_2013 %in% w_set,], pch=21, fill=NA, size=8, stroke=2, alpha=.6) +
ggtitle('English Properties (Lynott & Connell, 2009)') +
labs(x = "", y = "Varimax-rotated Principal Component 2") + theme_bw() +
theme( plot.background = element_blank(), panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.border = element_blank(),
axis.line = element_line(color = 'black'),
axis.title.x = element_text(colour = 'black', size = 14, margin=margin(7,0,0,0)),
axis.title.y = element_text(colour = 'black', size = 14, margin=margin(7,0,0,0)),
axis.text.x = element_text(size=9), axis.text.y = element_text(size=9),
plot.title = element_text(hjust = 0.5, size = 17, margin=margin(0,0,7,0)) ) +
geom_label_repel(data = eng_props[eng_props$English_Word_Lynott_Connell_2009_2013 %in% w_set,], aes(label = English_Word_Lynott_Connell_2009_2013), size = 6,
alpha = 0.77, color = 'black', box.padding = 1.5 )
# English Concepts
# check conditions for a PCA
# matrix
eng_conc = all[all$cat == 'Concept', c('English_Auditory_Lynott_Connell_2009_2013', 'English_Haptic_Lynott_Connell_2009_2013', 'English_Visual_Lynott_Connell_2009_2013')]
nrow(eng_conc)
eng_conc_matrix = cor(eng_conc, use = 'complete.obs')
eng_conc_matrix
round(eng_conc_matrix, 2)
# POOR: correlations not apt for a PCA, with too many below .3
# now on the raw data:
nrow(eng_conc)
cortest.bartlett(eng_conc)
# GOOD: Bartlett's test significant
# KMO: Kaiser-Meyer-Olkin Measure of Sampling Adequacy
KMO(eng_conc_matrix)
# Result: .48 = poor. PCA not strongly recommended. But we still do it
# because the purpose is graphical really.
# check determinant
det(eng_conc_matrix)
# GOOD: > 0.00001
# start off with unrotated PCA
pc1_eng_conc = psych::principal(eng_conc, nfactors = 3, rotate = "none")
pc1_eng_conc
# RESULT: Extract either one PC, acc to Kaiser's criterion, or two RCs, acc to
# Joliffe's (Field, Miles, & Field, 2012)
# Unrotated: scree plot
plot(pc1_eng_conc$values, type = "b")
# Result: two PCs obtain.
# Now with varimax rotation, Kaiser-normalized (by default):
# always preferable because it captures explained variance best.
pc2_eng_conc = psych::principal(eng_conc, nfactors = 2, rotate = "varimax",
scores = TRUE)
pc2_eng_conc
pc2_eng_conc$loadings
pc2_eng_conc$residual
pc2_eng_conc$fit
pc2_eng_conc$communality
# Results based on a Kaiser-normalizalized orthogonal (varimax) rotation
# (by default in psych::stats). Residuals bad: over 50% have absolute
# values > 0.05. Model fit good, > .90. Communalities good, all > .7.
# subset and add PCs
eng_concs = all[all$cat == 'Concept', ]
nrow(eng_concs)
eng_concs = cbind(eng_concs, pc2_eng_conc$scores)
summary(eng_concs$RC1, eng_concs$RC2)
eng_concs = eng_concs[eng_concs$normed == 'Dut_Eng' | eng_concs$normed ==
'English',]
nrow(eng_concs)
summary(eng_concs$RC1, eng_concs$RC2)
# Finally, plot
# Set sample words to show on plot (first word in each modality)
auditory_w = as.character(sort(eng_concs[eng_concs$English_Main_Lynott_Connell_2009_2013=='Auditory', 'English_Word_Lynott_Connell_2009_2013'])[1])
haptic_w = as.character(sort(eng_concs[eng_concs$English_Main_Lynott_Connell_2009_2013=='Haptic', 'English_Word_Lynott_Connell_2009_2013'])[1])
visual_w = as.character(sort(eng_concs[eng_concs$English_Main_Lynott_Connell_2009_2013=='Visual', 'English_Word_Lynott_Connell_2009_2013'])[1])
w_set = c(auditory_w, haptic_w, visual_w)
eng_concs$English_Main_Lynott_Connell_2009_2013 = recode(eng_concs$English_Main_Lynott_Connell_2009_2013, Auditory = "a", Haptic = "h", Visual = "v")
Engconcs = ggplot(eng_concs,
aes(RC1, RC2, label = as.character(English_Main_Lynott_Connell_2009_2013))) + stat_density2d (color = "gray87") +
geom_text(size = ifelse(eng_concs$English_Word_Lynott_Connell_2009_2013 %in% w_set, 8, 5),
fontface = ifelse(eng_concs$English_Word_Lynott_Connell_2009_2013 %in% w_set, 'bold', 'plain')) +
geom_point(data=eng_concs[eng_concs$English_Word_Lynott_Connell_2009_2013 %in% w_set,], pch=21, fill=NA, size=8, stroke=2, alpha=.6) +
ggtitle('English Concepts (Lynott & Connell, 2013)') +
labs(x = "Varimax-rotated Principal Component 1", y = "Varimax-rotated Principal Component 2") +
theme_bw() +
theme( plot.background = element_blank(), panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.border = element_blank(),
axis.line = element_line(color = 'black'),
axis.title.x = element_text(colour = 'black', size = 14, margin=margin(7,0,0,0)),
axis.title.y = element_text(colour = 'black', size = 14, margin=margin(7,0,0,0)),
axis.text.x = element_text(size=9), axis.text.y = element_text(size=9),
plot.title = element_text(hjust = 0.5, size = 17, margin=margin(0,0,7,0)) ) +
geom_label_repel(data = eng_concs[eng_concs$English_Word_Lynott_Connell_2009_2013 %in% w_set,], aes(label = English_Word_Lynott_Connell_2009_2013), size = 6,
alpha = 0.77, color = 'black', box.padding = 1.5 )
Engconcs # ! THE PLOT IS SHOWN BADLY ON HERE. PLEASE SEE THE SAVED PLOTS
# Now to save, run first line below and return to keep running. See your folder.
#png(file="Engconcs_highres.png", units="in", width=13, height=13, res=900)
#plot(Engconcs)
#dev.off()
# ON DUTCH NORMS
# Properties
# check conditions for a PCA
# matrix
Property = all[all$cat == 'Property' & !is.na(all$word), c('Auditory', 'Haptic', 'Visual')]
nrow(Property)
prop_matrix = cor(Property, use = 'complete.obs')
prop_matrix
round(prop_matrix, 2)
# POOR: correlations not apt for a PCA, with too many below .3
# now on the raw vars:
nrow(Property)
cortest.bartlett(Property)
# GOOD: Bartlett's test significant
# KMO: Kaiser-Meyer-Olkin Measure of Sampling Adequacy
KMO(prop_matrix)
# Result: .56 = mediocre. PCA not strongly recommended. But we still do it
# because the purpose is graphical only.
# check determinant
det(prop_matrix)
# GOOD: > 0.00001
# start off with unrotated PCA
pc1_prop = psych::principal(Property, nfactors = 3, rotate = "none")
pc1_prop
# RESULT: Only PC1, with eigenvalue > 1, should be extracted,
# acc to Kaiser's criterion (Jolliffe's threshold of 0.7 way too lax;
# Field, Miles, & Field, 2012)
# Unrotated: scree plot
plot(pc1_prop$values, type = "b")
# Result: one or two RCs should be extracted, converging with eigenvalues
# Now with varimax rotation, Kaiser-normalized (by default).
# Always preferable because it captures explained variance best.
# Compare eigenvalues w/ 1 & 2 factors
pc2_prop = psych::principal(Property, nfactors = 2, rotate = "varimax", scores = TRUE)
pc2_prop
pc2_prop$loadings
# good to extract 2 factors, as they both explain quite the same variance,
# and both surpass 1 eigenvalue
pc2_prop$residual
pc2_prop$fit
pc2_prop$communality
# Results based on a Kaiser-normalizalized orthogonal (varimax) rotation
# (by default in psych::stats). Residuals OK: fewer than 50% have absolute
# values > 0.05 (exactly 50% do).Model fit good, > .90.
# Communalities good, all > .7 (av = .83).
# subset and add PCs
props = all[all$cat == 'Property' & !is.na(all$word), ]
nrow(props)
props = cbind(props, pc2_prop$scores)
nrow(props)
# Finally, plot: letters+density (cf. Lynott & Connell, 2009, 2013)
# Set sample words to show on plot (first word in each modality)
auditory_w = as.character(sort(props[props$main=='Auditory', 'word'])[1])
haptic_w = as.character(sort(props[props$main=='Haptic', 'word'])[1])
visual_w = as.character(sort(props[props$main=='Visual', 'word'])[1])
w_set = c(auditory_w, haptic_w, visual_w)
props$main = recode(props$main, Auditory = "a", Haptic = "h", Visual = "v")
NLprops = ggplot(props,
aes(RC1, RC2, label = as.character(main))) + stat_density2d (color = "gray87") +
geom_text(size = ifelse(props$word %in% w_set, 12, 7),
fontface = ifelse(props$word %in% w_set, 'bold', 'plain')) +
geom_point(data=props[props$word %in% w_set,], pch=21, fill=NA, size=14, stroke=2, alpha=.6) +
ggtitle('Dutch Properties') +
labs(x = "Varimax-rotated Principal Component 1",
y = "Varimax-rotated Principal Component 2") + theme_bw() +
theme( plot.background = element_blank(), panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.border = element_blank(),
axis.line = element_line(color = 'black'),
axis.title.x = element_text(colour = 'black', size = 23, margin=margin(15,15,15,15)),
axis.title.y = element_text(colour = 'black', size = 23, margin=margin(15,15,15,15)),
axis.text.x = element_text(size=16), axis.text.y = element_text(size=16),
plot.title = element_text(hjust = 0.5, size = 32, face = "bold", margin=margin(15,15,15,15)),
plot.subtitle = element_text(hjust = 0.5, size = 20, margin=margin(2,15,15,15)) ) +
geom_label_repel(data = props[props$word %in% w_set,], aes(label = word), size = 8,
alpha = 0.77, color = 'black', box.padding = 1.5 )
NLprops # ! THE PLOT IS SHOWN BADLY ON HERE. PLEASE SEE THE SAVED PLOTS
# Now to save, run first line below and return to keep running. See your folder.
#png(file="NLprops_highres.png", units="in", width=13, height=13, res=900)
#plot(NLprops)
#dev.off()
# Adjust for combined plots:
NLprops2 = ggplot(props,
aes(RC1, RC2, label = as.character(main))) + stat_density2d (color = "gray87") +
geom_text(size = ifelse(props$word %in% w_set, 12, 7),
fontface = ifelse(props$word %in% w_set, 'bold', 'plain')) +
geom_point(data=props[props$word %in% w_set,], pch=21, fill=NA, size=14, stroke=2, alpha=.6) +
ggtitle('Dutch Properties') + labs(x = "", y = "") + theme_bw() +
theme( plot.background = element_blank(), panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.border = element_blank(),
axis.line = element_line(color = 'black'),
axis.title.x = element_text(colour = 'black', size = 23, margin=margin(15,15,15,15)),
axis.title.y = element_text(colour = 'black', size = 23, margin=margin(15,15,15,15)),
axis.text.x = element_text(size=16), axis.text.y = element_text(size=16),
plot.title = element_text(hjust = 0.5, size = 32, face = "bold", margin=margin(15,15,15,15)),
plot.subtitle = element_text(hjust = 0.5, size = 20, margin=margin(2,15,15,15)) ) +
geom_label_repel(data = props[props$word %in% w_set,], aes(label = word), size = 8,
alpha = 0.77, color = 'black', box.padding = 1.5 )
# Next: