-
Notifications
You must be signed in to change notification settings - Fork 1.3k
Expand file tree
/
Copy pathReport_s23.html
More file actions
2059 lines (1756 loc) · 68.1 KB
/
Report_s23.html
File metadata and controls
2059 lines (1756 loc) · 68.1 KB
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
<!DOCTYPE html>
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8"/>
<title>Derived Attributes and Dimensionality Reduction </title>
<script type="text/javascript">
window.onload = function() {
var imgs = document.getElementsByTagName('img'), i, img;
for (i = 0; i < imgs.length; i++) {
img = imgs[i];
// center an image if it is the only element of its parent
if (img.parentElement.childElementCount === 1)
img.parentElement.style.textAlign = 'center';
}
};
</script>
<!-- Styles for R syntax highlighter -->
<style type="text/css">
pre .operator,
pre .paren {
color: rgb(104, 118, 135)
}
pre .literal {
color: #990073
}
pre .number {
color: #099;
}
pre .comment {
color: #998;
font-style: italic
}
pre .keyword {
color: #900;
font-weight: bold
}
pre .identifier {
color: rgb(0, 0, 0);
}
pre .string {
color: #d14;
}
</style>
<!-- R syntax highlighter -->
<script type="text/javascript">
var hljs=new function(){function m(p){return p.replace(/&/gm,"&").replace(/</gm,"<")}function f(r,q,p){return RegExp(q,"m"+(r.cI?"i":"")+(p?"g":""))}function b(r){for(var p=0;p<r.childNodes.length;p++){var q=r.childNodes[p];if(q.nodeName=="CODE"){return q}if(!(q.nodeType==3&&q.nodeValue.match(/\s+/))){break}}}function h(t,s){var p="";for(var r=0;r<t.childNodes.length;r++){if(t.childNodes[r].nodeType==3){var q=t.childNodes[r].nodeValue;if(s){q=q.replace(/\n/g,"")}p+=q}else{if(t.childNodes[r].nodeName=="BR"){p+="\n"}else{p+=h(t.childNodes[r])}}}if(/MSIE [678]/.test(navigator.userAgent)){p=p.replace(/\r/g,"\n")}return p}function a(s){var r=s.className.split(/\s+/);r=r.concat(s.parentNode.className.split(/\s+/));for(var q=0;q<r.length;q++){var p=r[q].replace(/^language-/,"");if(e[p]){return p}}}function c(q){var p=[];(function(s,t){for(var r=0;r<s.childNodes.length;r++){if(s.childNodes[r].nodeType==3){t+=s.childNodes[r].nodeValue.length}else{if(s.childNodes[r].nodeName=="BR"){t+=1}else{if(s.childNodes[r].nodeType==1){p.push({event:"start",offset:t,node:s.childNodes[r]});t=arguments.callee(s.childNodes[r],t);p.push({event:"stop",offset:t,node:s.childNodes[r]})}}}}return t})(q,0);return p}function k(y,w,x){var q=0;var z="";var s=[];function u(){if(y.length&&w.length){if(y[0].offset!=w[0].offset){return(y[0].offset<w[0].offset)?y:w}else{return w[0].event=="start"?y:w}}else{return y.length?y:w}}function t(D){var A="<"+D.nodeName.toLowerCase();for(var B=0;B<D.attributes.length;B++){var C=D.attributes[B];A+=" "+C.nodeName.toLowerCase();if(C.value!==undefined&&C.value!==false&&C.value!==null){A+='="'+m(C.value)+'"'}}return A+">"}while(y.length||w.length){var v=u().splice(0,1)[0];z+=m(x.substr(q,v.offset-q));q=v.offset;if(v.event=="start"){z+=t(v.node);s.push(v.node)}else{if(v.event=="stop"){var p,r=s.length;do{r--;p=s[r];z+=("</"+p.nodeName.toLowerCase()+">")}while(p!=v.node);s.splice(r,1);while(r<s.length){z+=t(s[r]);r++}}}}return z+m(x.substr(q))}function j(){function q(x,y,v){if(x.compiled){return}var u;var s=[];if(x.k){x.lR=f(y,x.l||hljs.IR,true);for(var w in x.k){if(!x.k.hasOwnProperty(w)){continue}if(x.k[w] instanceof Object){u=x.k[w]}else{u=x.k;w="keyword"}for(var r in u){if(!u.hasOwnProperty(r)){continue}x.k[r]=[w,u[r]];s.push(r)}}}if(!v){if(x.bWK){x.b="\\b("+s.join("|")+")\\s"}x.bR=f(y,x.b?x.b:"\\B|\\b");if(!x.e&&!x.eW){x.e="\\B|\\b"}if(x.e){x.eR=f(y,x.e)}}if(x.i){x.iR=f(y,x.i)}if(x.r===undefined){x.r=1}if(!x.c){x.c=[]}x.compiled=true;for(var t=0;t<x.c.length;t++){if(x.c[t]=="self"){x.c[t]=x}q(x.c[t],y,false)}if(x.starts){q(x.starts,y,false)}}for(var p in e){if(!e.hasOwnProperty(p)){continue}q(e[p].dM,e[p],true)}}function d(B,C){if(!j.called){j();j.called=true}function q(r,M){for(var L=0;L<M.c.length;L++){if((M.c[L].bR.exec(r)||[null])[0]==r){return M.c[L]}}}function v(L,r){if(D[L].e&&D[L].eR.test(r)){return 1}if(D[L].eW){var M=v(L-1,r);return M?M+1:0}return 0}function w(r,L){return L.i&&L.iR.test(r)}function K(N,O){var M=[];for(var L=0;L<N.c.length;L++){M.push(N.c[L].b)}var r=D.length-1;do{if(D[r].e){M.push(D[r].e)}r--}while(D[r+1].eW);if(N.i){M.push(N.i)}return f(O,M.join("|"),true)}function p(M,L){var N=D[D.length-1];if(!N.t){N.t=K(N,E)}N.t.lastIndex=L;var r=N.t.exec(M);return r?[M.substr(L,r.index-L),r[0],false]:[M.substr(L),"",true]}function z(N,r){var L=E.cI?r[0].toLowerCase():r[0];var M=N.k[L];if(M&&M instanceof Array){return M}return false}function F(L,P){L=m(L);if(!P.k){return L}var r="";var O=0;P.lR.lastIndex=0;var M=P.lR.exec(L);while(M){r+=L.substr(O,M.index-O);var N=z(P,M);if(N){x+=N[1];r+='<span class="'+N[0]+'">'+M[0]+"</span>"}else{r+=M[0]}O=P.lR.lastIndex;M=P.lR.exec(L)}return r+L.substr(O,L.length-O)}function J(L,M){if(M.sL&&e[M.sL]){var r=d(M.sL,L);x+=r.keyword_count;return r.value}else{return F(L,M)}}function I(M,r){var L=M.cN?'<span class="'+M.cN+'">':"";if(M.rB){y+=L;M.buffer=""}else{if(M.eB){y+=m(r)+L;M.buffer=""}else{y+=L;M.buffer=r}}D.push(M);A+=M.r}function G(N,M,Q){var R=D[D.length-1];if(Q){y+=J(R.buffer+N,R);return false}var P=q(M,R);if(P){y+=J(R.buffer+N,R);I(P,M);return P.rB}var L=v(D.length-1,M);if(L){var O=R.cN?"</span>":"";if(R.rE){y+=J(R.buffer+N,R)+O}else{if(R.eE){y+=J(R.buffer+N,R)+O+m(M)}else{y+=J(R.buffer+N+M,R)+O}}while(L>1){O=D[D.length-2].cN?"</span>":"";y+=O;L--;D.length--}var r=D[D.length-1];D.length--;D[D.length-1].buffer="";if(r.starts){I(r.starts,"")}return R.rE}if(w(M,R)){throw"Illegal"}}var E=e[B];var D=[E.dM];var A=0;var x=0;var y="";try{var s,u=0;E.dM.buffer="";do{s=p(C,u);var t=G(s[0],s[1],s[2]);u+=s[0].length;if(!t){u+=s[1].length}}while(!s[2]);if(D.length>1){throw"Illegal"}return{r:A,keyword_count:x,value:y}}catch(H){if(H=="Illegal"){return{r:0,keyword_count:0,value:m(C)}}else{throw H}}}function g(t){var p={keyword_count:0,r:0,value:m(t)};var r=p;for(var q in e){if(!e.hasOwnProperty(q)){continue}var s=d(q,t);s.language=q;if(s.keyword_count+s.r>r.keyword_count+r.r){r=s}if(s.keyword_count+s.r>p.keyword_count+p.r){r=p;p=s}}if(r.language){p.second_best=r}return p}function i(r,q,p){if(q){r=r.replace(/^((<[^>]+>|\t)+)/gm,function(t,w,v,u){return w.replace(/\t/g,q)})}if(p){r=r.replace(/\n/g,"<br>")}return r}function n(t,w,r){var x=h(t,r);var v=a(t);var y,s;if(v){y=d(v,x)}else{return}var q=c(t);if(q.length){s=document.createElement("pre");s.innerHTML=y.value;y.value=k(q,c(s),x)}y.value=i(y.value,w,r);var u=t.className;if(!u.match("(\\s|^)(language-)?"+v+"(\\s|$)")){u=u?(u+" "+v):v}if(/MSIE [678]/.test(navigator.userAgent)&&t.tagName=="CODE"&&t.parentNode.tagName=="PRE"){s=t.parentNode;var p=document.createElement("div");p.innerHTML="<pre><code>"+y.value+"</code></pre>";t=p.firstChild.firstChild;p.firstChild.cN=s.cN;s.parentNode.replaceChild(p.firstChild,s)}else{t.innerHTML=y.value}t.className=u;t.result={language:v,kw:y.keyword_count,re:y.r};if(y.second_best){t.second_best={language:y.second_best.language,kw:y.second_best.keyword_count,re:y.second_best.r}}}function o(){if(o.called){return}o.called=true;var r=document.getElementsByTagName("pre");for(var p=0;p<r.length;p++){var q=b(r[p]);if(q){n(q,hljs.tabReplace)}}}function l(){if(window.addEventListener){window.addEventListener("DOMContentLoaded",o,false);window.addEventListener("load",o,false)}else{if(window.attachEvent){window.attachEvent("onload",o)}else{window.onload=o}}}var e={};this.LANGUAGES=e;this.highlight=d;this.highlightAuto=g;this.fixMarkup=i;this.highlightBlock=n;this.initHighlighting=o;this.initHighlightingOnLoad=l;this.IR="[a-zA-Z][a-zA-Z0-9_]*";this.UIR="[a-zA-Z_][a-zA-Z0-9_]*";this.NR="\\b\\d+(\\.\\d+)?";this.CNR="\\b(0[xX][a-fA-F0-9]+|(\\d+(\\.\\d*)?|\\.\\d+)([eE][-+]?\\d+)?)";this.BNR="\\b(0b[01]+)";this.RSR="!|!=|!==|%|%=|&|&&|&=|\\*|\\*=|\\+|\\+=|,|\\.|-|-=|/|/=|:|;|<|<<|<<=|<=|=|==|===|>|>=|>>|>>=|>>>|>>>=|\\?|\\[|\\{|\\(|\\^|\\^=|\\||\\|=|\\|\\||~";this.ER="(?![\\s\\S])";this.BE={b:"\\\\.",r:0};this.ASM={cN:"string",b:"'",e:"'",i:"\\n",c:[this.BE],r:0};this.QSM={cN:"string",b:'"',e:'"',i:"\\n",c:[this.BE],r:0};this.CLCM={cN:"comment",b:"//",e:"$"};this.CBLCLM={cN:"comment",b:"/\\*",e:"\\*/"};this.HCM={cN:"comment",b:"#",e:"$"};this.NM={cN:"number",b:this.NR,r:0};this.CNM={cN:"number",b:this.CNR,r:0};this.BNM={cN:"number",b:this.BNR,r:0};this.inherit=function(r,s){var p={};for(var q in r){p[q]=r[q]}if(s){for(var q in s){p[q]=s[q]}}return p}}();hljs.LANGUAGES.cpp=function(){var a={keyword:{"false":1,"int":1,"float":1,"while":1,"private":1,"char":1,"catch":1,"export":1,virtual:1,operator:2,sizeof:2,dynamic_cast:2,typedef:2,const_cast:2,"const":1,struct:1,"for":1,static_cast:2,union:1,namespace:1,unsigned:1,"long":1,"throw":1,"volatile":2,"static":1,"protected":1,bool:1,template:1,mutable:1,"if":1,"public":1,friend:2,"do":1,"return":1,"goto":1,auto:1,"void":2,"enum":1,"else":1,"break":1,"new":1,extern:1,using:1,"true":1,"class":1,asm:1,"case":1,typeid:1,"short":1,reinterpret_cast:2,"default":1,"double":1,register:1,explicit:1,signed:1,typename:1,"try":1,"this":1,"switch":1,"continue":1,wchar_t:1,inline:1,"delete":1,alignof:1,char16_t:1,char32_t:1,constexpr:1,decltype:1,noexcept:1,nullptr:1,static_assert:1,thread_local:1,restrict:1,_Bool:1,complex:1},built_in:{std:1,string:1,cin:1,cout:1,cerr:1,clog:1,stringstream:1,istringstream:1,ostringstream:1,auto_ptr:1,deque:1,list:1,queue:1,stack:1,vector:1,map:1,set:1,bitset:1,multiset:1,multimap:1,unordered_set:1,unordered_map:1,unordered_multiset:1,unordered_multimap:1,array:1,shared_ptr:1}};return{dM:{k:a,i:"</",c:[hljs.CLCM,hljs.CBLCLM,hljs.QSM,{cN:"string",b:"'\\\\?.",e:"'",i:"."},{cN:"number",b:"\\b(\\d+(\\.\\d*)?|\\.\\d+)(u|U|l|L|ul|UL|f|F)"},hljs.CNM,{cN:"preprocessor",b:"#",e:"$"},{cN:"stl_container",b:"\\b(deque|list|queue|stack|vector|map|set|bitset|multiset|multimap|unordered_map|unordered_set|unordered_multiset|unordered_multimap|array)\\s*<",e:">",k:a,r:10,c:["self"]}]}}}();hljs.LANGUAGES.r={dM:{c:[hljs.HCM,{cN:"number",b:"\\b0[xX][0-9a-fA-F]+[Li]?\\b",e:hljs.IMMEDIATE_RE,r:0},{cN:"number",b:"\\b\\d+(?:[eE][+\\-]?\\d*)?L\\b",e:hljs.IMMEDIATE_RE,r:0},{cN:"number",b:"\\b\\d+\\.(?!\\d)(?:i\\b)?",e:hljs.IMMEDIATE_RE,r:1},{cN:"number",b:"\\b\\d+(?:\\.\\d*)?(?:[eE][+\\-]?\\d*)?i?\\b",e:hljs.IMMEDIATE_RE,r:0},{cN:"number",b:"\\.\\d+(?:[eE][+\\-]?\\d*)?i?\\b",e:hljs.IMMEDIATE_RE,r:1},{cN:"keyword",b:"(?:tryCatch|library|setGeneric|setGroupGeneric)\\b",e:hljs.IMMEDIATE_RE,r:10},{cN:"keyword",b:"\\.\\.\\.",e:hljs.IMMEDIATE_RE,r:10},{cN:"keyword",b:"\\.\\.\\d+(?![\\w.])",e:hljs.IMMEDIATE_RE,r:10},{cN:"keyword",b:"\\b(?:function)",e:hljs.IMMEDIATE_RE,r:2},{cN:"keyword",b:"(?:if|in|break|next|repeat|else|for|return|switch|while|try|stop|warning|require|attach|detach|source|setMethod|setClass)\\b",e:hljs.IMMEDIATE_RE,r:1},{cN:"literal",b:"(?:NA|NA_integer_|NA_real_|NA_character_|NA_complex_)\\b",e:hljs.IMMEDIATE_RE,r:10},{cN:"literal",b:"(?:NULL|TRUE|FALSE|T|F|Inf|NaN)\\b",e:hljs.IMMEDIATE_RE,r:1},{cN:"identifier",b:"[a-zA-Z.][a-zA-Z0-9._]*\\b",e:hljs.IMMEDIATE_RE,r:0},{cN:"operator",b:"<\\-(?!\\s*\\d)",e:hljs.IMMEDIATE_RE,r:2},{cN:"operator",b:"\\->|<\\-",e:hljs.IMMEDIATE_RE,r:1},{cN:"operator",b:"%%|~",e:hljs.IMMEDIATE_RE},{cN:"operator",b:">=|<=|==|!=|\\|\\||&&|=|\\+|\\-|\\*|/|\\^|>|<|!|&|\\||\\$|:",e:hljs.IMMEDIATE_RE,r:0},{cN:"operator",b:"%",e:"%",i:"\\n",r:1},{cN:"identifier",b:"`",e:"`",r:0},{cN:"string",b:'"',e:'"',c:[hljs.BE],r:0},{cN:"string",b:"'",e:"'",c:[hljs.BE],r:0},{cN:"paren",b:"[[({\\])}]",e:hljs.IMMEDIATE_RE,r:0}]}};
hljs.initHighlightingOnLoad();
</script>
<style type="text/css">
body, td {
font-family: sans-serif;
background-color: white;
font-size: 13px;
}
body {
max-width: 800px;
margin: auto;
padding: 1em;
line-height: 20px;
}
tt, code, pre {
font-family: 'DejaVu Sans Mono', 'Droid Sans Mono', 'Lucida Console', Consolas, Monaco, monospace;
}
h1 {
font-size:2.2em;
}
h2 {
font-size:1.8em;
}
h3 {
font-size:1.4em;
}
h4 {
font-size:1.0em;
}
h5 {
font-size:0.9em;
}
h6 {
font-size:0.8em;
}
a:visited {
color: rgb(50%, 0%, 50%);
}
pre, img {
max-width: 100%;
}
pre {
overflow-x: auto;
}
pre code {
display: block; padding: 0.5em;
}
code {
font-size: 92%;
border: 1px solid #ccc;
}
code[class] {
background-color: #F8F8F8;
}
table, td, th {
border: none;
}
blockquote {
color:#666666;
margin:0;
padding-left: 1em;
border-left: 0.5em #EEE solid;
}
hr {
height: 0px;
border-bottom: none;
border-top-width: thin;
border-top-style: dotted;
border-top-color: #999999;
}
@media print {
* {
background: transparent !important;
color: black !important;
filter:none !important;
-ms-filter: none !important;
}
body {
font-size:12pt;
max-width:100%;
}
a, a:visited {
text-decoration: underline;
}
hr {
visibility: hidden;
page-break-before: always;
}
pre, blockquote {
padding-right: 1em;
page-break-inside: avoid;
}
tr, img {
page-break-inside: avoid;
}
img {
max-width: 100% !important;
}
@page :left {
margin: 15mm 20mm 15mm 10mm;
}
@page :right {
margin: 15mm 10mm 15mm 20mm;
}
p, h2, h3 {
orphans: 3; widows: 3;
}
h2, h3 {
page-break-after: avoid;
}
}
</style>
</head>
<body>
<p><link rel="stylesheet" href="http://netdna.bootstrapcdn.com/bootstrap/3.0.3/css/bootstrap.min.css"></p>
<style type="text/css"> body {padding: 10px 30px 10px 30px;} table,th, td {text-align: center;} </style>
<h1>Derived Attributes and Dimensionality Reduction </h1>
<p><strong>T. Evgeniou, INSEAD</strong></p>
<h2>What is this for?</h2>
<p>One of the key steps in Data Analytics is to generate meaningful attributes starting from possibly a large number of <strong>raw attributes</strong>. Consider for example the case of customer data, where for each customer we have a very large number of raw attributes (which could be, for example, <em>independent variables</em> in the case of a regression used to predict acquisition, purchases, or churn) ranging from demographic information, to what products they bought in the past, who their friends on various social media sites are, what websites they visit more often, how they rated a number of movies or products in general, whether they use their mobile phone mostly the weekends or in the mornings, how they responded to various surveys, etc. One can easily end up with tens if not thousands of such raw attributes, for thousands or millions of customers. In such cases it is virtually impossible to use some of the “advanced” methodologies developed for data analytics, or even simple ones such as linear regression. This is not only because of compuational reasons but, more important, because of statistical/mathematical reasons as, doing so entails a high risk that the estimated models (e.g. consumer behavior models) may be of very low quality in a statistical hence also practical sense. Moreover, the insights developed may not be practical or actionable as they may correspond to complicated statements involving a very large number of raw customer attributes. </p>
<p>In such situations, which arise almost always in practice, one needs to spend a lot of creative effort and time - based on a deep contextual knowledge - to generate new attributes (e.g. “this customer is price sensitive”, or “this customer likes comfort”, or “this customer is status conscious”, etc) from the original raw ones, which we call here <strong>derived attributes</strong>. These attributes can be generated manually, or using what we call <strong>data or dimensionality reduction statistical techniques</strong>. </p>
<p>We will consider a specific family of such statistical techniques which we will broadly call <strong>Factor Analysis techniques</strong>. Such techniques are often used at the early stages of a data analytics project, for example before running a regression with many independent variables (the raw attributes in that case), in order to summarize information (the variation) in correlated raw attributes using a smaller number of manageable <strong>factors</strong> - which are typically uncorrelated or independent. In the process one decreases the number of raw attributes while <strong>keeping most of the information in the data, in a statistical sense</strong>. </p>
<p>Such derived variables are usually useful also for managerial interpretation and action. For example, when analyzing survey data from students regarding the quality of a school, the quality of education may be a useful variable, but in a student survey one could instead ask several questions related to this: (1) Breadth of courses; (2) Depth of courses; (3) Quality of Instruction; (4) Practicality of coursework, etc. A “good linear combination” of these raw attributes may be more managerially useful in understanding student perceptions of the quality of education than each of these variables alone. If indeed these variables are highly related to the underlying construct of “quality of education”, Factor analysis will be able to summarize the information into such a single factor, while capturing most of the information in those “raw” attributes.</p>
<blockquote> <p>
Before proceeding on understanding the statistical techniques considered here, it is important to note that this is not the only approach to generating meaningful derived attributes from large numbers of raw ones: there is always “the manual approach”“ based on contextual knowledge and intuition, which can probably take a data analytics project already very far. However, in terms of mathematical techniques used in data analytics, factor analysis is one of the key ones when it comes to generating new meaningful derived attributes from the original raw ones.
</p> </blockquote>
<h2>Factor Analysis for Dimensionality Reduction using an Example</h2>
<p>There are many dimensionality reduction statistical methods. In this note we will go through the steps of one specific approach. We will do so using a simple example. For this example the "meaningful derived variables” will seem to be straightforward to design, which can help us with the intuition of what the method does. Having said this, before reading the analysis below, do try to think what “derived attributes” one could get from the raw ones below. You will see that even in this case it is not as obvious and you will most likely disagree with your colleagues about what the derived attributes should be: so we will let the numbers and statistics help us be more <em>objective and statistically correct</em>.</p>
<h3>The “Business Decision”</h3>
<p>We consider the core decision of an MBA admissions committee: <em>which applicants should we accept in the MBA program?</em> The school is interested in predicting the MBA participant's success in the future before offering admission.</p>
<h3>The Data</h3>
<p>To make this decision, the committee uses a number of data about the applicants. Let us consider for example the following attributes in evaluating an MBA application (of course in practice many more can be considered):</p>
<ol>
<li>GPA </li>
<li>GMAT score </li>
<li>Scholarships, fellowships won </li>
<li>Evidence of Communications skills (debating competition, personal interview score) </li>
<li>Prior Job Experience </li>
<li>Organizational Experience </li>
<li>Other extra curricular achievements </li>
</ol>
<p>Let us assume that this data is converted into a numerical scale from 1-7. For example: a numerical rating may be given to the fellowships based on the prestige and the number of fellowships won. Job experience may be rated on the number of years on the job, with a numerical weighting for the level of the job in the managerial ladder.</p>
<p>This is how the first 20 data looks:
<br></p>
<div class="row">
<div class="col-md-6">
<!-- Table generated in R 3.2.3 by googleVis 0.5.10 package -->
<!-- Thu Jan 28 12:00:11 2016 -->
<!-- jsHeader -->
<script type="text/javascript">
// jsData
function gvisDataTableID1fb0786711f1 () {
var data = new google.visualization.DataTable();
var datajson =
[
[
"1",
3,
580,
2,
3.5,
5,
3.8,
4
],
[
"2",
3.2,
570,
2,
3.8,
6,
3.8,
3.8
],
[
"3",
3.7,
690,
3,
3.3,
3,
3.2,
3.6
],
[
"4",
3.9,
760,
3,
3.8,
5,
3.9,
3.2
],
[
"5",
2.8,
480,
2,
3.2,
6,
3.8,
3.8
],
[
"6",
3.4,
520,
2.5,
2.6,
2,
2.5,
2.4
],
[
"7",
3.6,
670,
3,
3.7,
4,
3.5,
2.9
],
[
"8",
3.6,
760,
3,
3.9,
5,
3.3,
3.2
],
[
"9",
2.8,
380,
1,
2,
3,
2.9,
3.1
],
[
"10",
3.6,
560,
3,
2.8,
2,
1,
2.8
],
[
"11",
2.8,
480,
2,
3.2,
4,
3.4,
3.7
],
[
"12",
3.6,
590,
3.4,
3.4,
5,
3.7,
3.6
],
[
"13",
3.8,
720,
3.8,
3.8,
6,
3.8,
3.7
],
[
"14",
3.5,
680,
3,
3.6,
6,
3.4,
3.8
],
[
"15",
3.9,
700,
3.7,
3.8,
2,
2.1,
2.7
],
[
"16",
3.3,
560,
2.8,
3.4,
5,
3.8,
2.7
],
[
"17",
2.8,
440,
1,
3.3,
5,
3.5,
3.8
],
[
"18",
2.5,
430,
1,
3.1,
3,
3.1,
2.9
],
[
"19",
2.5,
420,
1,
2.8,
2,
2.3,
2.5
],
[
"20",
3.8,
680,
2.8,
3.7,
6,
3.2,
3.8
]
];
data.addColumn('string','Variables');
data.addColumn('number','GPA');
data.addColumn('number','GMAT');
data.addColumn('number','Fellow');
data.addColumn('number','Comm');
data.addColumn('number','Job.Ex');
data.addColumn('number','Organze');
data.addColumn('number','Extra');
data.addRows(datajson);
return(data);
}
// jsDrawChart
function drawChartTableID1fb0786711f1() {
var data = gvisDataTableID1fb0786711f1();
var options = {};
options["allowHtml"] = true;
options["showRowNumber"] = true;
options["width"] = 1220;
options["height"] = 400;
options["allowHTML"] = true;
options["page"] = "disable";
var chart = new google.visualization.Table(
document.getElementById('TableID1fb0786711f1')
);
chart.draw(data,options);
}
// jsDisplayChart
(function() {
var pkgs = window.__gvisPackages = window.__gvisPackages || [];
var callbacks = window.__gvisCallbacks = window.__gvisCallbacks || [];
var chartid = "table";
// Manually see if chartid is in pkgs (not all browsers support Array.indexOf)
var i, newPackage = true;
for (i = 0; newPackage && i < pkgs.length; i++) {
if (pkgs[i] === chartid)
newPackage = false;
}
if (newPackage)
pkgs.push(chartid);
// Add the drawChart function to the global list of callbacks
callbacks.push(drawChartTableID1fb0786711f1);
})();
function displayChartTableID1fb0786711f1() {
var pkgs = window.__gvisPackages = window.__gvisPackages || [];
var callbacks = window.__gvisCallbacks = window.__gvisCallbacks || [];
window.clearTimeout(window.__gvisLoad);
// The timeout is set to 100 because otherwise the container div we are
// targeting might not be part of the document yet
window.__gvisLoad = setTimeout(function() {
var pkgCount = pkgs.length;
google.load("visualization", "1", { packages:pkgs, callback: function() {
if (pkgCount != pkgs.length) {
// Race condition where another setTimeout call snuck in after us; if
// that call added a package, we must not shift its callback
return;
}
while (callbacks.length > 0)
callbacks.shift()();
} });
}, 100);
}
// jsFooter
</script>
<!-- jsChart -->
<script type="text/javascript" src="https://www.google.com/jsapi?callback=displayChartTableID1fb0786711f1"></script>
<!-- divChart -->
<div id="TableID1fb0786711f1"
style="width: 1220; height: 400;">
</div>
<p></div>
</div>
<br> <br></p>
<p>We will see some descriptive statistics of the data later, when we get into statistical analysis.</p>
<h3>The Approach</h3>
<p>How can this data inform the school's admission decisions? Does this data capture some “derived attributes” that may have a meaning? If you were to derive 2 or 3 attrributes from the data above (by combining them in various ways, for example), what would those be? Which raw attributes would be “linked” with the derived ones you think of? <strong>Try it</strong> intuitively before reading any further…</p>
<p>Intuitively it may seem that the data above capture two fundamental abilities that affect the success of students in their management careers: </p>
<ol>
<li>Basic intelligence </li>
<li>Team and Leadership skills. </li>
</ol>
<p>The school may be interested for example in picking students who score high on these two areas. In this case, of course the admissions commitee in theory could just ask the applicants two questions: </p>
<ol>
<li>“How intelligent are you?” </li>
<li>“How strong are your team and leadership skills?” </li>
</ol>
<p>As you can imagine, asking these questions would not only make the admissions interviewers look naive, but would also lead to very noisy and misleading answers: of course everyone will just answer both questions with the highest mark. So instead of asking these “naive” questions, the school is using <em>raw attributes/data</em> like the ones above, which can also be gathered easier. The idea then is to see how this data can be “translated” in meaningful derived attributes that, for example, could capture the “equivalent answers” one could get if one were to ask directly the two naive questions above - or possibly other such “complex”“ questions. </p>
<blockquote> <p>
Factor analysis is a statistical approach for finding a few "hidden” derived attributes in data by combining together groups of the original raw attributes in such a way that the least information in the original data is lost - in a statistical sense. It is part of a general class of statistical methodologies used to do what is often called “dimensionality reduction”. </p> </blockquote>
<p>Back to our example, if there is some way in which we could reduce the 7 attributes into a smaller set of, say, 2 or 3 attributes, then we can reduce the data to a more understandable form so that our decision making process can be made potentially simpler, more actionable, and easier to interpret and justify - without losing much information in the original data. It is much easier to make tradeoffs between two or three attributes than it is between 10 or 20 attributes (look at any survey or application form and you will see that there are easily more than 20 questions). Hence, </p>
<blockquote> <p>
Data reduction is a very useful step in helping us interpret the data and make decisions.
</p> </blockquote>
<p>Like for our example, theory may suggest that there are really one or two basic factors (like intelligence and leadership skills) that lead to success in a management career. The various attributes are really different manifestations of these basic factors. But maybe there are other hidden derived variables (factors) in the data we have: instead of us manually combining raw attributes into meaningful derived ones, which not only is diffucult with many data but also dangerous as we impose our biases, let's get <em>factor analysis</em> to do the job for us - and use our intuition and judgment in the process. </p>
<p>Let's now see <strong>a</strong> process for using factor analysis in order to create derived attributes, the goal of this report. </p>
<h3>A Process for Dimensionality Reduction</h3>
<blockquote> <p>
It is important to remember that Data Analytics Projects require a delicate balance between experimentation, intuition, but also following (once a while) a process to avoid getting fooled by randomness and “finding results and patterns” that are mainly driven by our own biases and not by the facts/data themselves.
</p> </blockquote>
<p>There is <em>not one</em> process for factor analysis. However, we have to start somewhere, so we will use the following process:</p>
<h4>Factor Analysis in 6 steps</h4>
<ol>
<li><p>Confirm the data in metric </p></li>
<li><p>Decide whether to scale or standardize the data</p></li>
<li><p>Check the correlation matrix to see if Factor Analysis makes sense</p></li>
<li><p>Develop a scree plot and decide on the number of factors to be derived</p></li>
<li><p>Interpret the factors (consider factor rotations - technical but useful)</p></li>
<li><p>Save factor scores for subsequent analyses</p></li>
</ol>
<p>Let's follow these steps.</p>
<h4>Step 1: Confirm the data are metric</h4>
<p>Steps 1-3 are about specific descriptive characteristics of the data. In particular, the methods we consider in this note require that the data are <em>metric</em> (step 1): this means not only that all data are numbers, but also that the numbers have an actual numerical meaning, that is 1 is less than 2 which is less than 3 etc. If we have other types of data (e.g. gender, categories that are not comparable, etc), there are other methods to use. However, for now we will only consider a specific method, which we will also mis-use for non-numeric data for simplicity. </p>
<p>The data we use here have the following descriptive statistics: </p>
<div class="row">
<div class="col-md-6">
<!-- Table generated in R 3.2.3 by googleVis 0.5.10 package -->
<!-- Thu Jan 28 12:00:12 2016 -->
<!-- jsHeader -->
<script type="text/javascript">
// jsData
function gvisDataTableID1fb03c023a55 () {
var data = new google.visualization.DataTable();
var datajson =
[
[
"GPA",
2.5,
2.8,
3.45,
3.31,
3.62,
3.9,
0.47
],
[
"GMAT",
380,
480,
575,
583.5,
682.5,
760,
119.44
],
[
"Fellow",
1,
2,
2.8,
2.45,
3,
3.8,
0.91
],
[
"Comm",
2,
3.18,
3.4,
3.34,
3.73,
3.9,
0.49
],
[
"Job.Ex",
2,
3,
5,
4.25,
5.25,
6,
1.52
],
[
"Organze",
1,
3.05,
3.4,
3.2,
3.8,
3.9,
0.73
],
[
"Extra",
2.4,
2.88,
3.4,
3.3,
3.8,
4,
0.52
]
];
data.addColumn('string','Variables');
data.addColumn('number','min');
data.addColumn('number','X25.percent');
data.addColumn('number','median');
data.addColumn('number','mean');
data.addColumn('number','X75.percent');
data.addColumn('number','max');
data.addColumn('number','std');
data.addRows(datajson);
return(data);
}
// jsDrawChart
function drawChartTableID1fb03c023a55() {
var data = gvisDataTableID1fb03c023a55();
var options = {};
options["allowHtml"] = true;
options["showRowNumber"] = true;
options["width"] = 1220;
options["height"] = 216;
options["allowHTML"] = true;
options["page"] = "disable";
var chart = new google.visualization.Table(
document.getElementById('TableID1fb03c023a55')
);
chart.draw(data,options);
}
// jsDisplayChart
(function() {
var pkgs = window.__gvisPackages = window.__gvisPackages || [];
var callbacks = window.__gvisCallbacks = window.__gvisCallbacks || [];
var chartid = "table";
// Manually see if chartid is in pkgs (not all browsers support Array.indexOf)
var i, newPackage = true;
for (i = 0; newPackage && i < pkgs.length; i++) {
if (pkgs[i] === chartid)
newPackage = false;
}
if (newPackage)
pkgs.push(chartid);
// Add the drawChart function to the global list of callbacks
callbacks.push(drawChartTableID1fb03c023a55);
})();
function displayChartTableID1fb03c023a55() {
var pkgs = window.__gvisPackages = window.__gvisPackages || [];
var callbacks = window.__gvisCallbacks = window.__gvisCallbacks || [];
window.clearTimeout(window.__gvisLoad);
// The timeout is set to 100 because otherwise the container div we are
// targeting might not be part of the document yet
window.__gvisLoad = setTimeout(function() {
var pkgCount = pkgs.length;
google.load("visualization", "1", { packages:pkgs, callback: function() {
if (pkgCount != pkgs.length) {
// Race condition where another setTimeout call snuck in after us; if
// that call added a package, we must not shift its callback
return;
}
while (callbacks.length > 0)
callbacks.shift()();
} });
}, 100);
}
// jsFooter
</script>
<!-- jsChart -->
<script type="text/javascript" src="https://www.google.com/jsapi?callback=displayChartTableID1fb03c023a55"></script>
<!-- divChart -->
<div id="TableID1fb03c023a55"
style="width: 1220; height: 216;">
</div>
<p></div>
</div></p>
<blockquote> <p>
Note that one should spend a lot of time getting a feeling of the data based on simple summary statistics and visualizations: good data analytics require that we understand our data very well.
</p> </blockquote>
<h4>Step 2: Decide whether to scale or standardize the data</h4>
<p>Note that for this data, while 6 of the “survey” data are on a similar scale, namely 1-7, there is one variable that is about 2 orders of magnitude larger: the GMAT variable. Having some variables with a very different range/scale can often create problems: <strong>most of the “results” may be driven by a few large values</strong>, more so that we would like. To avoid such issues, one has to consider whether or not to <strong>standardize the data</strong> by making some of the initial raw attributes have, for example, mean 0 and standard deviation 1 (e.g. scaledGMAT = (GMAT-mean(GMAT))/sd(GMAT) ), or scaling them between 0 and 1 (e.g. scaledGMAT=(GMAT-min(GMAT))/(max(GMAT)-min(GMAT))). Here is for example the R code for the first approach, if we want to standardize all attributes:</p>
<pre><code class="r">ProjectDatafactor_scaled=apply(ProjectDataFactor,2, function(r) {if (sd(r)!=0) res=(r-mean(r))/sd(r) else res=0*r; res})
</code></pre>
<p>Notice now the summary statistics of the scaled dataset:</p>
<p><br></p>
<div class="row">
<div class="col-md-6">
<!-- Table generated in R 3.2.3 by googleVis 0.5.10 package -->
<!-- Thu Jan 28 12:00:12 2016 -->
<!-- jsHeader -->
<script type="text/javascript">
// jsData
function gvisDataTableID1fb010eb7f () {
var data = new google.visualization.DataTable();
var datajson =
[
[
"GPA",
-1.72,
-1.08,
0.31,
0,
0.68,
1.27,
1
],
[
"GMAT",
-1.7,
-0.87,
-0.07,
0,
0.83,
1.48,
1
],
[
"Fellow",
-1.6,
-0.5,
0.39,
0,
0.61,
1.49,
1
],
[
"Comm",
-2.73,
-0.33,
0.13,
0,
0.8,
1.16,
1
],
[
"Job.Ex",
-1.48,
-0.82,
0.49,
0,
0.66,
1.15,
1
],
[
"Organze",
-2.99,
-0.2,
0.27,
0,
0.82,
0.95,
1
],
[
"Extra",
-1.75,
-0.83,
0.19,
0,
0.97,
1.36,
1
]
];
data.addColumn('string','Variables');
data.addColumn('number','min');
data.addColumn('number','X25.percent');
data.addColumn('number','median');
data.addColumn('number','mean');
data.addColumn('number','X75.percent');
data.addColumn('number','max');
data.addColumn('number','std');
data.addRows(datajson);
return(data);
}
// jsDrawChart
function drawChartTableID1fb010eb7f() {
var data = gvisDataTableID1fb010eb7f();
var options = {};
options["allowHtml"] = true;
options["showRowNumber"] = true;
options["width"] = 1220;
options["height"] = 216;
options["allowHTML"] = true;
options["page"] = "disable";
var chart = new google.visualization.Table(
document.getElementById('TableID1fb010eb7f')
);
chart.draw(data,options);
}
// jsDisplayChart
(function() {
var pkgs = window.__gvisPackages = window.__gvisPackages || [];
var callbacks = window.__gvisCallbacks = window.__gvisCallbacks || [];
var chartid = "table";
// Manually see if chartid is in pkgs (not all browsers support Array.indexOf)
var i, newPackage = true;
for (i = 0; newPackage && i < pkgs.length; i++) {
if (pkgs[i] === chartid)
newPackage = false;
}
if (newPackage)
pkgs.push(chartid);
// Add the drawChart function to the global list of callbacks
callbacks.push(drawChartTableID1fb010eb7f);
})();
function displayChartTableID1fb010eb7f() {
var pkgs = window.__gvisPackages = window.__gvisPackages || [];
var callbacks = window.__gvisCallbacks = window.__gvisCallbacks || [];
window.clearTimeout(window.__gvisLoad);
// The timeout is set to 100 because otherwise the container div we are
// targeting might not be part of the document yet
window.__gvisLoad = setTimeout(function() {
var pkgCount = pkgs.length;
google.load("visualization", "1", { packages:pkgs, callback: function() {
if (pkgCount != pkgs.length) {
// Race condition where another setTimeout call snuck in after us; if
// that call added a package, we must not shift its callback
return;
}
while (callbacks.length > 0)
callbacks.shift()();
} });
}, 100);
}
// jsFooter
</script>
<!-- jsChart -->
<script type="text/javascript" src="https://www.google.com/jsapi?callback=displayChartTableID1fb010eb7f"></script>
<!-- divChart -->
<div id="TableID1fb010eb7f"
style="width: 1220; height: 216;">
</div>
<p></div>
</div></p>
<p><br>
As expected all variables have mean 0 and standard deviation 1. </p>
<p>While this is typically a necessary step, one has to always do it with care: some times you may want your analytics findings to be driven mainly by a few attributes that take large values; other times having attributes with different scales may imply something about those attributes. For example, when students rate their schools on various factors on a 1-7 scale, if the variability is minimal on a certain variable (e.g. satisfaction about the IT infrastructure of the school) but very high on another one (e.g. satisfaction with job placement), then standardization will reduce the real big differences in placement satisfaction and magnify the small differences in IT infrastructure satisfaction. In many such cases one may choose to skip step 2 for some of the raw attributes. Hence standardization is not a necessary data transformation step, and you should use it judiciously. </p>
<h4>Step 3: Check correlation matrix to see if Factor Analysis makes sense</h4>
<p>The type of dimensionality reduction methods we will use here “groups together raw attributes that are highly correlated”. Other methods (there are many!) use different criteria to create derived variables. For this to be feasible, it is necessary that the original raw attributes do have large enough correlations (e.g. more than 0.5 in absolute value, or simply statistically significant). It is therefore useful to see the correlation matrix of the original attributes - something that one should anyway always do in order to develop a better understanding of the data. </p>
<p>This is the correlation matrix of the 7 original variable we use for factor analysis (Note: this would be the same for the standardized ones if the standardization is done as above; there is a mathematical reason for this that we will not explore - you could confirm it yourself):</p>
<table border="1" class="data table table-bordered table-condensed">
<thead class="thead">
<tr>
<th></th>
<th>GPA</th>
<th>GMAT</th>
<th>Fellow</th>
<th>Comm</th>
<th>Job.Ex</th>
<th>Organze</th>
<th>Extra</th>
</tr>
</thead>
<tbody>
<tr>
<td align="right">GPA</td>
<td align="right" style="background-color: rgb(0,255,0)">1.00</td>
<td align="right" style="background-color: rgb(25,255,25)">0.90</td>
<td align="right" style="background-color: rgb(20,255,20)">0.92</td>