forked from tevgeniou/BuybacksIssuers
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlib_helpers.R
1212 lines (1081 loc) · 61.4 KB
/
lib_helpers.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
# Copyright 2013, Satrapade
# by T. Evgeniou, V. Kapartzianis, N. Nassuphis and D. Spinellis
# Dual licensed under the MIT or GPL Version 2 licenses.
# 11/2015 functions car_table and calendar_table added by T. Evgeniou and E. Junqué de Fortuny
# Required R libraries (need to be installed - it can take a few minutes the first time you run the project)
if (ifelse(!exists("run_shiny_tool"), T, run_shiny_tool == 0)) { # When deploying a shiny app on the shiny server we should not load all R libraries
# installs all necessary libraries from CRAN
get_libraries <- function(filenames_list) {
lapply(filenames_list,function(thelibrary){
if (do.call(require,list(thelibrary)) == FALSE)
do.call(install.packages,list(thelibrary))
do.call(library,list(thelibrary))
})
}
libraries_used=c("stringr","gtools","foreign","reshape2","digest","timeDate","devtools","knitr","graphics",
"grDevices","xtable","sqldf","stargazer","data.table","shiny","htmlwidgets",
"Hmisc","vegan","fpc","GPArotation","FactoMineR","cluster",
"psych","stringr","googleVis", "png","ggplot2","googleVis", "gridExtra","RcppArmadillo","xts","DescTools")
get_libraries(libraries_used)
}
Rcpp::sourceCpp('lib_helpers.cpp', embeddedR=FALSE)
options(stringsAsFactors=FALSE)
# Summary table
my_summary <- function(x){
nas = sum(is.na(x))
x=x[!is.na(x)]
res = c(mean(x), median(x),min(x),quantile(x,0.1), quantile(x,0.25), quantile(x,0,95), quantile(x,0.9), max(x), sd(x), length(x))
names(res) <- c("mean", "p50", "min", "p10", "p25", "p75","p90","max","sd", "N")
res
}
non_na_mean <- function(x) { mean(x[!is.na(x) & x != 0]) }
non_na_sd <- function(x) { sd(x[!is.na(x) & x != 0]) }
remove_initialization_time <- function(x,min_date=NULL) { # Note: Added also min_date as an argument!! THis way we remove data from any given date!
if (!is.null(min_date)){
res = x[as.Date(names(x)) >= as.Date(min_date)]
} else {
if (x[1] != 0){
res = x
} else {
if (sum(x!=0) != 0){
#res = tail(res, -head(which(diff(res)!=0),1))
#names(res) <- tail(names(res), -head(which(diff(res)!=0),1))
tmp <- names(x)
ending <- -head(which(diff(x)!=0),1)
res = tail(x, ending)
names(res) <- tail(tmp, ending)
}
}
}
res
}
yeardays<-function(x){
fd<-as.integer(as.Date(head(names(x),1),format="%Y-%m-%d"))
ld<-as.integer(as.Date(tail(names(x),1),format="%Y-%m-%d"))
365.25*length(x)/(ld-fd)
}
vol_pa<-function(x,exclude_zero=(x!=0), holidays = holidayNYSE()) {
if(is.null(names(x))){ good_days <- TRUE }else { good_days <- isBizday(as.timeDate(names(x)), holidays, wday = 1:5) }
x1<-drop(x[exclude_zero & good_days])
sqrt(yeardays(x1))*sd(x1)
}
sharpe<-function(x,exclude_zero=(x!=0), holidays = holidayNYSE() ) {
if(is.null(names(x))){ good_days <- TRUE }else { good_days <- isBizday(as.timeDate(names(x)), holidays, wday = 1:5) }
x1<-drop(x[exclude_zero & good_days])
round(yeardays(x1)*mean(x1)/vol_pa(x1),digits=2) # annualized
}
bps<-function(x,exclude_zero=(x!=0))round(10000*mean(drop(x[exclude_zero])),digits=2)
drawdown<-function(x)round(100*max(cummax(cumsum(x))-cumsum(x)),digits=2)
pnl_stats<-function(x, show_tr = FALSE, show_gr = FALSE){
if(class(x)=="matrix")if(ncol(x)>1)x<-x[,1]
ret<-c(Ret=round(100*yeardays(x)*mean(x),digits=1),Vol=round(100*vol_pa(x),digits=1),Sharpe=round(sharpe(x),2),DD=round(drawdown(x),1))
ret
}
pnl_plot<-function(x,...){
pargs<-as.list(match.call(expand.dots=TRUE))
if(!"ylab" %in% names(pargs)) ylab<-deparse(substitute(x)) else ylab<-pargs$ylab
if(!"main" %in% names(pargs)) main<-paste(names(pnl_stats(x)),pnl_stats(x),sep=":",collapse=" ") else main<-pargs$main
plot_arguments<-c(list(x=cumsum(x*100),type="l",ylab=ylab,cex.main = 0.9, main=main,axes=FALSE),pargs[setdiff(names(pargs),c("","x","ylab","main"))])
do.call(plot,plot_arguments)
if(!is.null(names(x))){
axis(1,at=seq(1,length(x),length.out=5),labels=names(x)[seq(1,length(x),length.out=5)])
axis(2)
} else { axis(1); axis(2)}
}
pnl_matrix<-function(perf, digits = 2){
month_map<-c("01"="Jan","02"="Feb","03"="Mar","04"="Apr","05"="May","06"="Jun","07"="Jul","08"="Aug","09"="Sep","10"="Oct","11"="Nov","12"="Dec")
perf_dates<-structure(do.call(rbind,strsplit(names(perf),"-")),dimnames=list(names(perf),c("Year","Month","Day")))
perf_dates[,"Month"]<-month_map[perf_dates[,"Month"]]
perf_years <- sort(unique(perf_dates[,"Year"]))
perf_months <- month_map
res<-structure(
outer(perf_years,perf_months,function(i_vec,j_vec)mapply(function(i,j){
perf_ndx <- perf_dates[,"Year"]==i & perf_dates[,"Month"]==j
if(sum(perf_ndx)==0)return(NA)
prod(perf[perf_ndx]+1)-1
},i_vec,j_vec)),
dimnames=list(perf_years,perf_months)
)
round(cbind(res,Year=apply(res,1,function(r)prod(r[!is.na(r)]+1)-1))*100,digits=2)
}
show_stats<-function(x)paste(names(pnl_stats(x)),pnl_stats(x),sep=":",collapse=" \\textbar ")
###
# if number of non-zeroes of x is less than n, return 0, else return the mean of the non-zero entries
non_zero_mean<-function(x,n=1)ifelse(sum(scrub(x)!=0)<n,0,mean(scrub(x)[scrub(x)!=0]))
# moving average of previous n elements : 0 for first n-1 elements
ma<-function(x,n,f=identity){res<-as.numeric(filter(f(x),rep(1/n,n),method="convolution",sides=1,circular=FALSE)); ifelse(is.na(res),0,res)}
# moving sum of previous n elements : 0 for first n-1 elements
ms<-function(x,n,f=identity){res<-as.numeric(filter(f(x),rep(1,n),method="convolution",sides=1,circular=FALSE)); ifelse(is.na(res),0,res)}
# shift forward if n +ve , backward if n -ve.
moving_average<-function(n)function(x)ma(x,n)
moving_sum<-function(n)function(x)ms(x,n)
shift<-function(a,n=1,filler=0){
x<-switch(class(a),matrix=a,matrix(a,ncol=1,dimnames=list(names(a),NULL)))
if(n==0)return(x)
if(n>0){
rbind(matrix(filler,ncol=ncol(x),nrow=n),head(x,-n))
} else {
rbind(tail(x,n),matrix(filler,ncol=ncol(x),nrow=abs(n)))
}
}
# replace non-finite elements in x with zeroes
scrub<-function(x){
if(length(x)==0)return(0)
x[which(!is.finite(x))]<-0
x
}
# Fill NAs with earlier values
fill_NA_previous <- function(x, lastfill = F){
if (sum(!is.na(x)) == 0)
return(x)
non_na = which(!is.na(x))
if (length(non_na) == 1)
return(c(rep(x[non_na], non_na-1), x[non_na:length(x)]))
if (!(1 %in% non_na)) non_na = c(1,non_na)
x = c(unlist(lapply(1:(length(non_na)-1), function(i) rep(x[non_na[i]],non_na[i+1] - non_na[i]))), x[tail(non_na,1):length(x)])
if (lastfill)
x[tail(non_na,1):length(x)] <- x[tail(non_na,1)]
x
}
######
# roll function over multiple matrix rows, in parallel. mapply for matrices.
multiroll_fun<-function(fun,w,...,MoreArgs=NULL,verbose=identity){
data<-list(...)
if(length(data)==0)stop("multiroll: data required")
if(length(data)!=length(formals(fun)))stop("multiroll: fun arg count differs from data length")
if(!all(unlist(lapply(data,class))=="matrix"))stop("multiroll: only matrix")
if( length(unique(unlist(lapply(data,nrow))))!=1 )stop("multiroll: all matrices must have same rows")
p<-nrow(data[[1]])
ndx<-t(matrix(c(0,w-1),nrow=2,ncol=p-w+1)+t(matrix(1:(p-w+1),ncol=2,nrow=p-w+1)))
res<-t(simplify2array(apply(ndx,1,function(r){
window_res<-do.call(fun,c(lapply(data,function(m)m[r[1]:r[2],,drop=FALSE]),MoreArgs))
if(class(verbose)=="function")cat(r[2]," ",verbose(window_res),"\n")
window_res
})))
if(class(res)=="matrix")rownames(res)<-rownames(data[[1]])[ndx[,2]] else names(res)<-rownames(data[[1]])[ndx[,2]]
res
}
# apply function to matrix, matrix result
row_apply<-function(m,f,...){
if(class(m)!="matrix")return(NULL)
if(!any(class(f)%in%c("function","standardGeneric")))return(NULL)
mcol<-ncol(m)
res<-t(apply(m,1,function(r){
row_res<-drop(unlist(f(r,...)))
if(length(row_res)==1)return(rep(row_res,mcol))
if(length(row_res)==mcol)return(row_res)
return(rep(0,length(r)))
}))
dimnames(res)<-dimnames(m)
res
}
# apply function to matrix, matrix result
col_apply<-function(m,f,...){
if(class(m)!="matrix")return(NULL)
if(!any(class(f)%in%c("function","standardGeneric")))return(NULL)
mrow<-nrow(m)
res<-apply(m,2,function(r){
row_res<-drop(unlist(f(r,...)))
if(length(row_res)==1)return(rep(row_res,mrow))
if(length(row_res)==mrow)return(row_res)
return(rep(0,length(r)))
})
dimnames(res)<-dimnames(m)
res
}
"%-%"<-row_apply
"%|%"<-col_apply
################################################################################################################
################################################################################################################
# data is a daysXsecurities matrix of dayly returns.
# the %|% operator applies a function column-wise
# moving_average(window) returns a function that does it.
rolling_variance<-function(data,window){
average_square <- (data*data)%|%moving_average(window)
average_move <- data%|%moving_average(window)
average_square - average_move*average_move
}
# Deals with 0s and NAs:
rolling_variance_nonzero<-function(data,window){
data2 = (data*data)
average_non_zero <- (data!=0)%|%moving_sum(window)
average_square <- data2%|%moving_sum(window)
average_move <- data%|%moving_sum(window)
average_square = ifelse(average_non_zero, average_square/average_non_zero, 0)
average_move = ifelse(average_non_zero, average_move/average_non_zero, 0)
average_square - average_move*average_move
}
rolling_covariance<-function(security,hedge,window){
comovement <- (security*hedge)%|%moving_average(window)
product_of_means <- (security%|%moving_average(window))*(hedge%|%moving_average(window))
comovement-product_of_means
}
rolling_correlation<-function(security1,security2,window){
covariance<-rolling_covariance(security1,security2,window)
variance1<-rolling_variance(security1,window)
variance2<-rolling_variance(security2,window)
ifelse(variance1>0&variance2>0,covariance/sqrt(variance1*variance2),0)
}
rolling_beta<-function(security,hedge,window){
variance<-rolling_variance(hedge,window)
covariance<-rolling_covariance(security,hedge,window)
ifelse(variance>0,covariance/variance,0)
}
alpha_ff <- function(ri,RiskFactors,Risk_Factors_Equation) {
data = RiskFactors
data$ri = ri
form = as.formula(Risk_Factors_Equation)
model = lm(form,data=data)
alpha = summary(model)$coefficients[1]
}
################################################################################################################
################################################################################################################
# DATASET CREATION RELATED FUNCTIONS
################################################################################################################
################################################################################################################
# A helper that creates yearly matrices for each firm characteristic,
# aligned with the monthly data. Note the fiscal year use
# (datadate is the end of the fiscal year)
create_yearly_data <- function(value_used, template_matrix,all_compustat_data){
tmp_data = as.data.frame(dcast(all_compustat_data, datadate ~ LPERMNO,
fun.aggregate = function(r) ifelse(length(unique(r)) > 1, NA, unique(r)),
value.var=value_used))
# Convert tmp_data to matrix with dates as rownames
tmp = as.character(tmp_data$datadate)
tmp_data$datadate <- NULL
tmp_data <- as.matrix(tmp_data)
rownames(tmp_data) <- tmp
# Align with monthly data now
rownames(tmp_data) <- rownames(
template_matrix)[match(
str_sub(rownames(tmp_data),start=1,end=7),
str_sub(rownames(template_matrix),start=1,end=7))]
tmp_data = tmp_data[,intersect(colnames(tmp_data),
colnames(template_matrix))]
res = NA*template_matrix
res[rownames(tmp_data),colnames(tmp_data)] <- tmp_data
# Fill in the gaps
# we need to start using the data from the "next month",
# after the fiscal year end
res = apply(res,2,function(r) {
x = fill_NA_previous(r);
if(is.na(tail(x,1)) & sum(!is.na(x))) {
x[pmin(length(x),tail(which(!is.na(x)),1):
(tail(which(!is.na(x)),1)+11))] <- x[tail(which(!is.na(x)),1)]
}
c(NA,head(x,-1))
})
rownames(res) <- rownames(template_matrix)
res
}
# Creates cross-sectional percentile based scores for a given company characteristic.
get_cross_section_score <- function(company_feature_matrix, company_feature_matrix_used=NULL, zero_special = F){
datacol = ncol(company_feature_matrix)
data_used = company_feature_matrix
if (!is.null(company_feature_matrix_used))
data_used = cbind(company_feature_matrix,company_feature_matrix_used)
tmp = t(apply(data_used,1,function(r){
r_scored = r[1:datacol]
r_ecdf = r_scored
if (!is.null(company_feature_matrix_used))
r_ecdf = r[(datacol+1):length(r)]
res = r_scored*NA
if (!zero_special){
r_scored = scrub(r_scored)
r_ecdf = scrub(r_ecdf)
if (sum(r_ecdf !=0)) { # Note: we don't use both NAs and 0s
score_fun = ecdf(r_ecdf[r_ecdf!=0])
res = ifelse(r_scored!=0, score_fun(r_scored),NA)
}
} else {
if (sum(!is.na(r_ecdf))) { # Note: we don't use NAs only here
score_fun = ecdf(r_ecdf[!is.na(r_ecdf)])
res = ifelse(!is.na(r_scored), score_fun(r_scored),NA)
}
}
res
}))
rownames(tmp) <- rownames(company_feature_matrix)
tmp
}
################################################################################################################
################################################################################################################
# Simple analyses
################################################################################################################
################################################################################################################
# Returns the alpha, based on a running regression
# Riskfactors could be any (date x factor) vector or data frame and will be used in an lm:
# lm("return ~ alpha + beta x factors")
# You can also specify a special "RF" factor in the df which will automatically go in the left-hand side:
# lm("(return - RF) ~ factors)s
alpha_lm <- function(ri,Riskfactors,hedge_days, trade = 0) {
#data check
if (class(Riskfactors) != "data.frame") {
dates <- names(Riskfactors)
Riskfactors = data.frame(market=Riskfactors)
rownames(Riskfactors) <- dates
}
#create the right formula
RFfield <- which(colnames(Riskfactors) == "RF")
NotRFfield <- colnames(Riskfactors)[setdiff(1:ncol(Riskfactors),RFfield)]
if(length(RFfield) == 1) {
form <- as.formula(paste("(rj - RF) ~ ",paste(NotRFfield,collapse = "+")))
} else {
form <- as.formula(paste("rj ~ ",paste(NotRFfield,collapse = "+")))
}
#do a running coefficient estimation
runcoeff <- function(rj) {
data = Riskfactors[rownames(Riskfactors) >= min(names(rj)) & rownames(Riskfactors) <= max(names(rj)),,drop=F]
data$rj = rj
model = fastLm(form,data=data)
return(summary(model)$coefficients[,1])
}
coeff <- running(ri,fun=runcoeff,width=hedge_days,allow.fewer = T)
coeffcorrection <- shift(t(coeff[NotRFfield,]),trade) * Riskfactors[,NotRFfield]
if(dim(coeffcorrection)[1] !=1)
coeffcorrection <- apply(coeffcorrection,1,sum)
alpha <- (ri - coeffcorrection)
if(length(RFfield) == 1)
alpha <- alpha - Riskfactors[,RFfield]
alpha <- alpha * (coeffcorrection!=0)*(ri!=0)
return(alpha)
}
beta_expost <- function(ri,Riskfactors) {
#data check
if (class(Riskfactors) != "data.frame") {
dates <- names(Riskfactors)
Riskfactors = data.frame(market=Riskfactors)
rownames(Riskfactors) <- dates
}
useonly = intersect(names(ri), rownames(Riskfactors))
Riskfactors = Riskfactors[useonly,]
ri = ri[useonly]
#create the right formula
RFfield <- which(colnames(Riskfactors) == "RF")
NotRFfield <- colnames(Riskfactors)[setdiff(1:ncol(Riskfactors),RFfield)]
if(length(RFfield) == 1) {
form <- as.formula(paste("(ri - RF) ~ ",paste(NotRFfield,collapse = "+")))
} else {
form <- as.formula(paste("ri ~ ",paste(NotRFfield,collapse = "+")))
}
data = Riskfactors
data$ri = ri
model = fastLm(form,data=data)
return(summary(model)$coefficients)
}
################################################################################################################
################################################################################################################
# Econometric methods
################################################################################################################
################################################################################################################
################################################################################################################
#Builds an IRATS table. Returns need to be monthly,
# same for the risk factors. See description of method in the table captions of http://tevgeniou.github.io/BuybacksIssuers/
################################################################################################################
car_table <- function(returns,Event.Date,Risk_Factors_Monthly,min_window = -6, max_window = 48,formula_used="(ri - RF) ~ Delta + SMB + HML + RMW + CMA",value.weights = 1) {
#data check
if (class(Risk_Factors_Monthly) != "data.frame") {
dates <- names(Risk_Factors_Monthly)
Risk_Factors_Monthly = data.frame(market=Risk_Factors_Monthly)
rownames(Risk_Factors_Monthly) <- dates
}
# Make all data have monthly dates on first of month:
rownames(returns) <- paste(str_sub(rownames(Risk_Factors_Monthly), start=1,end=7),"01",sep="-")
rownames(Risk_Factors_Monthly) <- paste(str_sub(rownames(Risk_Factors_Monthly), start=1,end=7),"01",sep="-")
###
factors_used = setdiff(unlist(str_split(gsub("~", ",", gsub("\\-", ",", gsub("\\+", ",", gsub("\\)", "",gsub("\\(", "",formula_used))))), " , ")),"ri")
min_window = min(1,min_window)
max_window = max(-1,max_window)
allmonths = min_window:max_window
if (sum(!(factors_used %in% colnames(Risk_Factors_Monthly))))
stop(paste("car_table misses the risk factors: ",factors_used[!(factors_used %in% colnames(Risk_Factors_Monthly))]))
factors_used_noRF = setdiff(factors_used, "RF")
if (ncol(returns) < length(factors_used_noRF) + 1){
results = matrix(0,nrow=length(allmonths), ncol = 3)
results = rbind(results, rep(ncol(returns)))
betas = matrix(0,nrow=length(allmonths), ncol = length(factors_used_noRF))
betasstderr = matrix(0,nrow=length(allmonths), ncol = length(factors_used_noRF))
colnames(results) <- c("CAL","t-stat","p-value")
rownames(results) <- c(ifelse(allmonths > 0, paste("+",allmonths,sep=""), allmonths),"Observations")
colnames(results) <- c("CAR","t-stat","p-value")
colnames(betas) <- factors_used_noRF
colnames(betasstderr) <- factors_used_noRF
all_results = list(results = results, betas = betas, betasstderr = betasstderr)
return(all_results)
}
#Step 1: Build an event matrix, where all the events are aligned by event month as opposed to calendar month
Event.Date = paste(str_sub(Event.Date,start=1,end=7), "01", sep="-") # just in case
Row.Date <- as.Date(rownames(returns))
Row.Date_number = as.numeric(Row.Date)
Event.Date_number = as.numeric(as.Date(Event.Date))
firstday = as.numeric(as.Date(head(rownames(returns),1)))
lastday = as.numeric(as.Date(tail(rownames(returns),1)))
firstHit = match(Event.Date_number,Row.Date_number)
form = as.formula(formula_used)
Risk_Factors_Monthly = Risk_Factors_Monthly[,factors_used]
if (!("RF" %in% factors_used)){
Risk_Factors_Monthly = cbind(Risk_Factors_Monthly,matrix(0,nrow=nrow(Risk_Factors_Monthly)))
colnames(Risk_Factors_Monthly)[ncol(Risk_Factors_Monthly)] <- "RF"
}
#Build event_month x company x factor matrix
EVENT_ALIGNED <- array(0, c(length(allmonths), ncol(returns), ncol(Risk_Factors_Monthly)+1))
starting = pmax(1,pmin(nrow(returns),firstHit + min_window))
ending = pmax(1,pmin(nrow(returns),firstHit + max_window))
Risk_Factors_Matrix <- as.matrix(Risk_Factors_Monthly)
# THIS IS THE SLOW PART
for(ev in 1:length(Event.Date)) {
EVENT_ALIGNED[(1 + (starting[ev] - (firstHit[ev] + min_window))):(length(allmonths) - (firstHit[ev] + max_window-ending[ev])),ev,] <- cbind(Risk_Factors_Matrix[starting[ev]:ending[ev],], returns[starting[ev]:ending[ev],ev])
}
dimnames(EVENT_ALIGNED) <- list( ifelse(allmonths > 0, paste("+",allmonths,sep=""), allmonths),
c(paste(colnames(returns),Event.Date_number)),
c(colnames(Risk_Factors_Monthly),"ri"))
#Step2: now for each month, calculate the CAR. need to include month 0 and set that one to 0
alphas <- rep(0,length(allmonths))
betas <- array(0,c(length(allmonths),ncol(Risk_Factors_Monthly)-1))
betasstderr <- array(0,c(length(allmonths),ncol(Risk_Factors_Monthly)-1))
stderr <- rep(0,length(allmonths))
dfs <- as.integer(rep(0,length(allmonths)))
event_alphas = matrix(0,nrow = length(Event.Date_number), ncol = length(allmonths))
match_ini = match(Event.Date_number, Row.Date_number)
for (i in 1:length(allmonths)) {
if (allmonths[i] !=0) { #we do not consider the month of the event
hitnow = match_ini + allmonths[i]
ret <- EVENT_ALIGNED[i,,]
non_zeros = which(ret[,"ri"]!=0 & (hitnow <= length(Row.Date_number)) & (hitnow >= 1))
ret <- ret[non_zeros,] # WE NEED THIS HERE!!!!!
if (nrow(ret) > ncol(ret)){
model = fastLm(form,data=data.frame(ret,row.names = NULL))
model.summary = summary(model)
alphas[i] = model.summary$coefficients[1,"Estimate"]
stderr[i] = model.summary$coefficients[1, "StdErr"]
the_betas = model.summary$coefficients[2:nrow(model.summary$coefficients), "Estimate"]
betas[i,] <- the_betas
betasstderr[i,] <- model.summary$coefficients[2:nrow(model.summary$coefficients), "StdErr"]
dfs[i] = df.residual(model)
event_alphas[non_zeros,i] <- ret[,"ri"] - ret[,"RF"] - ret[,names(the_betas)]%*%matrix(the_betas,ncol=1)
} else{
alphas[i] = 0
betas[i,] <- 0
betasstderr[i,] <- 0
stderr[i] = 0
dfs[i] = 0
event_alphas[non_zeros,i] <- 0
}
}
}
event_alphas_aggregate = 0*event_alphas
#summary CAR output: aggregate results for all windows
results <- array(0,c(length(allmonths)+1,3))
if (0 %in% allmonths){
for (i in 1:(which(allmonths==0)-1)){
thealpha = sum(alphas[(which(allmonths==0)-1):i]) # this is i:1 since we want to end up with a 0 at event month - 1 (and at event month)
thestderr = sqrt(sum(stderr[(which(allmonths==0)-1):i]*stderr[(which(allmonths==0)-1):i])) # this is 1:i since we want to add them all from the oldest month
tstat = ifelse(thestderr, thealpha / thestderr, 0)
results[i,] <- c(
thealpha,
tstat,
2 * pt(abs(tstat), df = dfs[i], lower.tail = FALSE) #pvalue
)
}
for (i in (which(allmonths==0)+1):length(allmonths)){
thealpha = sum(alphas[(which(allmonths==0)+1):i])
thestderr = sqrt(sum(stderr[(which(allmonths==0)+1):i]*stderr[(which(allmonths==0)+1):i]))
tstat = ifelse(thestderr, thealpha / thestderr, 0)
results[i,] <- c(
thealpha,
tstat,
2 * pt(abs(tstat), df = dfs[i], lower.tail = FALSE) #pvalue
)
}
} else {
for (i in 1:length(allmonths)){
thealpha = sum(alphas[1:i])
thestderr = sqrt(sum(stderr[1:i]*stderr[1:i]))
tstat = ifelse(thestderr, thealpha / thestderr, 0)
results[i,] <- c(
thealpha,
tstat,
2 * pt(abs(tstat), df = dfs[i], lower.tail = FALSE) #pvalue
)
}
}
results[,1] = 100*results[,1]
rownames(results) <- c(ifelse(allmonths > 0, paste("+",allmonths,sep=""), allmonths),"Observations")
colnames(results) <- c("CAR","t-stat","p-value")
results[nrow(results),] <- rep(length(Event.Date_number),ncol(results))
colnames(betas) <- factors_used_noRF
colnames(betasstderr) <- factors_used_noRF
all_results = list(results = results, betas = betas, betasstderr = betasstderr,event_alphas = event_alphas,dfs = dfs)
return(all_results)
}
################################################################################################################
# Builds a Calendar Time table. Returns need to be monthly,
# same for the risk factors. See description of method in the table captions of http://tevgeniou.github.io/BuybacksIssuers/
################################################################################################################
calendar_table <- function(returns,Event.Date, Risk_Factors_Monthly,min_window = -6, max_window = 48,formula_used="(ri - RF) ~ Delta + SMB + HML + RMW + CMA",value.weights = 1) {
#data check
if (class(Risk_Factors_Monthly) != "data.frame") {
dates <- names(Risk_Factors_Monthly)
Risk_Factors_Monthly = data.frame(market=Risk_Factors_Monthly)
rownames(Risk_Factors_Monthly) <- dates
}
# Make all data have monthly dates on first of month:
rownames(returns) <- paste(str_sub(rownames(Risk_Factors_Monthly), start=1,end=7),"01",sep="-")
rownames(Risk_Factors_Monthly) <- paste(str_sub(rownames(Risk_Factors_Monthly), start=1,end=7),"01",sep="-")
###
factors_used = setdiff(unlist(str_split(gsub("~", ",", gsub("\\-", ",", gsub("\\+", ",", gsub("\\)", "",gsub("\\(", "",formula_used))))), " , ")),"ri")
factors_used_noRF = setdiff(factors_used, "RF")
min_window = min(1,min_window)
max_window = max(-1,max_window)
allmonths = min_window:max_window
if (sum(!(factors_used %in% colnames(Risk_Factors_Monthly))))
stop(paste("calendar_table misses the risk factors: ",factors_used[!(factors_used %in% colnames(Risk_Factors_Monthly))]))
if (ncol(returns) < length(factors_used_noRF) + 1){
results = matrix(0,nrow=length(allmonths), ncol = 3)
results = rbind(results, rep(ncol(returns),3))
betas = matrix(0,nrow=length(allmonths), ncol = length(factors_used_noRF))
betasstderr = matrix(0,nrow=length(allmonths), ncol = length(factors_used_noRF))
dfs <- as.integer(rep(0,length(allmonths)))
colnames(results) <- c("CAL","t-stat","p-value")
rownames(results) <- c(ifelse(allmonths > 0, paste("+",allmonths,sep=""), allmonths),"Observations")
colnames(results) <- c("CAR","t-stat","p-value")
colnames(betas) <- factors_used_noRF
colnames(betasstderr) <- factors_used_noRF
all_results = list(results = results, betas = betas, betasstderr = betasstderr)
return(all_results)
}
Row.Date <- as.Date(rownames(returns))
form = as.formula(formula_used)
Event.Date = paste(str_sub(Event.Date,start=1,end=7), "01", sep="-") # just in case
Risk_Factors_Monthly = Risk_Factors_Monthly[,factors_used]
if (!("RF" %in% factors_used)){
Risk_Factors_Monthly = cbind(Risk_Factors_Monthly,matrix(0,nrow=nrow(Risk_Factors_Monthly)))
colnames(Risk_Factors_Monthly)[ncol(Risk_Factors_Monthly)] <- "RF"
}
alphas <- rep(0,length(allmonths))
betas <- array(0,c(length(allmonths),ncol(Risk_Factors_Monthly)-1))
betasstderr <- array(0,c(length(allmonths),ncol(Risk_Factors_Monthly)-1))
stderr <- rep(0,length(allmonths))
dfs <- as.integer(rep(0,length(allmonths)))
Event.Date_number = as.numeric(as.Date(Event.Date))
Row.Date_number = as.numeric(Row.Date)
#Now create aggregate returns based on the window and calculate the model
results <- array(0,c(length(allmonths)+1,3))
max_index = length(Row.Date_number)
start_match = match(Event.Date_number,Row.Date_number)
for (i in 1:length(allmonths)) {
w = allmonths[i]
hitnow = Row.Date_number[pmax(1,pmin(start_match+w,max_index))]
ret <- returns
if (w > 0) ret <- calendar_table_helper1(ret, Row.Date_number, Event.Date_number, hitnow)
#for(j in 1:length(Event.Date))
# ret[ Row.Date_number <= Event.Date_number[j] | Row.Date_number > hitnow[j],j ] <- 0
if (w < 0) ret <- calendar_table_helper2(ret, Row.Date_number, Event.Date_number, hitnow)
# for(j in 1:length(Event.Date))
# ret[ Row.Date_number >= Event.Date_number[j] | Row.Date_number < hitnow[j],j ] <- 0
ri <- row_weights(ret, value.weights)
if (sum(ri!=0) > 10) { ### For special cases...
ri <- ri[ head(which(ri !=0),1) : tail(which(ri!=0),1)]
data = Risk_Factors_Monthly
data = data[rownames(data) >= min(names(ri)) & rownames(data) <= max(names(ri)),]
data$ri = ri
form = as.formula(form)
model = fastLm(form,data=data)
results[i,] <- c(
summary(model)$coefficients[1,"Estimate"]*100,
summary(model)$coefficients[1,"t.value"],
summary(model)$coefficients[1,"p.value"]
)
betas[i,] <- summary(model)$coefficients[2:nrow(summary(model)$coefficients), "Estimate"]
betasstderr[i,] <- summary(model)$coefficients[2:nrow(summary(model)$coefficients), "StdErr"]
dfs[i] = df.residual(model)
}
}
rownames(results) <- c(ifelse(allmonths > 0, paste("+",allmonths,sep=""), allmonths),"Observations")
colnames(results) <- c("CAL","t-stat","p-value")
results[nrow(results),] <- rep(length(Event.Date),ncol(results))
colnames(betas) <- factors_used_noRF
colnames(betasstderr) <- factors_used_noRF
all_results = list(results = results, betas = betas, betasstderr = betasstderr,dfs=dfs)
return(all_results)
}
####################################################################################
#Builds a stock specific regression like Brennan, Chordia and Subrahmanyam (1998)
####################################################################################
# Split in two parts as the first one is more generic
# PART I OF BSC1998_event_study: This is a general function that gets the betas and actual returns for a series of months for the events
event_study_factor_coeffs <- function(returns,Event.Date, Risk_Factors_Monthly,formula_used="(ri - RF) ~ Delta + SMB + HML + RMW + CMA", rolling_window=60, timeperiods_requested = 1:48, min_data = 10){
# assumes returns has one column per event, so number of columns equal to Event.Date. Rows are months
# Keep track also of the risk and stock returns the next month - where we will do the predictions
factors_used = setdiff(unlist(str_split(gsub("~", ",", gsub("\\-", ",", gsub("\\+", ",", gsub("\\)", "",gsub("\\(", "",formula_used))))), " , ")),"ri")
if (sum(!(factors_used %in% colnames(Risk_Factors_Monthly))))
stop(paste("BSC1998_event_study misses the risk factors: ",factors_used[!(factors_used %in% colnames(Risk_Factors_Monthly))]))
factors_used_noRF = setdiff(factors_used, "RF")
number_of_factors = length(factors_used_noRF)
form = as.formula(formula_used)
Risk_Factors_Monthly = Risk_Factors_Monthly[,factors_used]
if (!("RF" %in% factors_used)){
Risk_Factors_Monthly = cbind(Risk_Factors_Monthly,matrix(0,nrow=nrow(Risk_Factors_Monthly)))
colnames(Risk_Factors_Monthly)[ncol(Risk_Factors_Monthly)] <- "RF"
}
#Step 1: Build an event matrix, where all the events are aligned by event month as opposed to calendar month
returns_month = str_sub(rownames(returns), start=1,end=7)
Event.Date = str_sub(Event.Date,start=1,end=7)
firstHit = match(Event.Date,returns_month)
if (sum(is.na(firstHit)) !=0)
print("\nSOME EVENTS SEEM TO BE OUTISDE THE RANGE OF DATES OF THE RISK FACTORS...\n")
res_init = structure(rep(NA,number_of_factors + (number_of_factors+1) + 1), .Names = c(factors_used_noRF, paste(factors_used,"ret", sep="_"), "actual_ret"))
max_time = length(returns_month)
## Step 1: Estimate Factor Loadings for the requested months (month 0 is the event month)
lapply(1:length(Event.Date), function(i){
if (i%%50 == 1) cat(i,",")
stock_returns = structure(returns[,i], .Names = returns_month)
event_row = firstHit[i]
stock_factors = t(Reduce(cbind,lapply(timeperiods_requested, function(themonth){
#cat(themonth,",")
res = res_init
period_used = min(max_time, max(1,event_row+themonth - rolling_window)):min(max_time, max(1,event_row+themonth -1)) # We use -1 but can also use -2 to use till two months before
period_to_predict = min(max_time, max(1,event_row+themonth))
if (length(period_used) > min_data & tail(period_used,1) < period_to_predict){ # in case we are at the last month available for this stocks
ret = cbind(Risk_Factors_Monthly,stock_returns)[period_used,] # Ignore NAs and 0s
colnames(ret)[ncol(ret)]<- "ri"
if (sum(!is.na(ret[,"ri"])) >= min_data){
ret <- ret[!is.na(ret[,"ri"]),]
model = fastLm(form,data=data.frame(ret,row.names = NULL))
thebetas = summary(model)$coefficients[2:nrow(summary(model)$coefficients), "Estimate"]
res = c(thebetas,as.numeric(Risk_Factors_Monthly[period_to_predict,]), stock_returns[period_to_predict])
names(res) <- c(factors_used_noRF, paste(factors_used,"ret", sep="_"), "actual_ret")
}
}
res
})))
rownames(stock_factors) <- timeperiods_requested
stock_factors
})
}
# PART II OF BSC1998_event_study
BSC1998_event_study_coeffs <- function(Estimated_returns,company_features,timeperiods_requested = 1:48,square_features=NULL){
# assumes returns has one column per event, so number of columns equal to Event.Date. Rows are months
# Assumes factor_loadings comes from event_study_factor_coeffs using the exact same inputs
## Step 1: Estimate Factor Loadings for the requested months (so month 0 is the event month)
# Assumes these are inputs: It is done using the function event_study_factor_coeffs above
## Step 2: Calculate Monthly Estimated Risk-adjusted Return (Given as input)
## Step 3: Run Cross-Section Regression in Each Post-Event Month, from 1-48 months
features_to_square = intersect(colnames(company_features),square_features)
colnames_used = colnames(company_features)
if (length(features_to_square) !=0)
colnames_used = c(colnames_used,paste(features_to_square,"square",sep=" "))
C_mt_coefficients = Reduce(cbind,lapply(1:ncol(Estimated_returns), function(month){
month_returns = Estimated_returns[,month]
useonly = which(!is.na(month_returns) & !is.na(apply(company_features,1,sum)))
res = rep(NA, ncol(company_features)+1)
if (length(useonly) > ncol(company_features) + 1){
company_features_used = company_features[useonly,,drop=F]
# square and demean what is needed
if (!is.null(features_to_square)){
for (square_i in features_to_square){
company_features_used[,square_i] <- company_features_used[,square_i] - mean(company_features_used[,square_i])
company_features_used = cbind(company_features_used,company_features_used[,square_i]*company_features_used[,square_i])
}
}
cross_sectional_data = cbind(company_features_used, month_returns[useonly])
colnames(cross_sectional_data) <- c(paste("Ind", 1:ncol(company_features_used), sep=""), "ret")
cross_setional_form = as.formula(paste("ret", str_c(paste("Ind", 1:ncol(company_features_used), sep=""), collapse=" + "), sep=" ~ "))
model = fastLm(cross_setional_form,data=data.frame(cross_sectional_data,row.names = NULL))
res = summary(model)$coefficients[,1]
}
res
}))
colnames(C_mt_coefficients) <- timeperiods_requested
if (!is.null(colnames(company_features)))
rownames(C_mt_coefficients) <- c("Intercept", colnames_used)
C_mt_coefficients
## Step 4: Aggregate C_mt_coefficients over 48 Post-Event Months: Time-Series Average of C_mt_coefficients (This can be done outside, for whatever months one needs)
}
## Step 4: Aggregate C_mt_coefficients over 48 Post-Event Months: Time-Series Average of C_mt_coefficients (This can be done outside, for whatever months one needs)
BSC1998_coeffs_tmp_aggregator <- function(BSC1998_coefficients,the_months_needed = c("12", "24", "36","48")){
BSC1998_coeffs = t(Reduce(cbind,lapply(the_months_needed, function(i){
apply(BSC1998_coefficients,1,function(r){
useonly = which(colnames(BSC1998_coefficients)=="1"):which(colnames(BSC1998_coefficients)==i)
mean(r[useonly])
})
})))
rownames(BSC1998_coeffs) <- paste("month", the_months_needed)
BSC1998_tstats = t(Reduce(cbind,lapply(the_months_needed, function(i){
apply(BSC1998_coefficients,1,function(r){
useonly = which(colnames(BSC1998_coefficients)=="1"):which(colnames(BSC1998_coefficients)==i)
mean(r[useonly])/(sd(r[useonly])/sqrt(length(useonly)))
})
})))
rownames(BSC1998_tstats) <- rep("t-stat", nrow(BSC1998_tstats))
BSC1998_pvalue = t(Reduce(cbind,lapply(the_months_needed, function(i){
apply(BSC1998_coefficients,1,function(r){
useonly = which(colnames(BSC1998_coefficients)=="1"):which(colnames(BSC1998_coefficients)==i)
x = mean(r[useonly])/(sd(r[useonly])/sqrt(length(useonly)))
ifelse(x > 0, 1-pt(x,df=length(useonly-1)), pt(x,df=length(useonly-1)))
})
})))
rownames(BSC1998_pvalue) <- rep("p-value", nrow(BSC1998_tstats))
BSC1998<- Reduce(rbind,lapply(1:nrow(BSC1998_coeffs), function(i) {res = rbind(100*BSC1998_coeffs[i,],BSC1998_tstats[i,],BSC1998_pvalue[i,]); rownames(res)<-c(rownames(BSC1998_coeffs)[i], rownames(BSC1998_tstats)[i],rownames(BSC1998_pvalue)[i]); res}))
colnames(BSC1998) <- rownames(BSC1998_coefficients)
t(BSC1998)
}
################################################################################################################
#####################################################################################
### ADDITIONAL KEY CODE FROM BUYBACKS/ISSUERS PROJECT
#####################################################################################
################################################################################################################
#Normal performance Function
Performance <- function(date_to_start, date_to_end, Dates, returns_used, event){
Start_Date <- Dates[which(rownames(Dates) == date_to_start),]
End_Date <- Dates[which(rownames(Dates) == date_to_end),]
if (event)
res = Reduce(cbind,lapply(1:length(Start_Date), function(i) 252*non_zero_mean(returns_used[rownames(returns_used) >= Start_Date[i] & rownames(returns_used) <= End_Date[i],i])))
if (!event)
res = Reduce(cbind,lapply(1:length(End_Date), function(i) 252*non_zero_mean(returns_used[(rownames(returns_used) > End_Date[i] | rownames(returns_used) < Start_Date[i]),i])))
res
}
#PNL Matrix function (for any subset of companies to use)
PNL_matrix_BB <- function(date_to_start, date_to_end, company_subset, Dates, returns_used, event=1){
returns_used <- returns_used[,company_subset, drop=F]
Start_Date <- Dates[which(rownames(Dates) == date_to_start),company_subset]
End_Date <- Dates[which(rownames(Dates) == date_to_end),company_subset]
dates_used = rownames(returns_used)
if (event){
dates_matrix= mapply(function(s,e){
dates_used >= s & dates_used <= e
},Start_Date, End_Date)
res = returns_used*dates_matrix
}
if (!event){
dates_matrix = Reduce(cbind,lapply(1:length(Start_Date), function(i) (dates_used > End_Date[i])))
res = returns_used*dates_matrix
}
res
}
Betas <- function(x,y,Dates, EVENT_RETURNS,Riskfactors){
if (class(Riskfactors) != "matrix")
Riskfactors = matrix(Riskfactors,ncol=1)
Start_Date <- Dates[which(rownames(Dates) == x),]
End_Date <- Dates[which(rownames(Dates) == y),]
non_zeros <- which(apply(Riskfactors!=0,1,sum)!=0)
Riskfactors <- Riskfactors[non_zeros,,drop=F]
tmp_matrix <- EVENT_RETURNS[non_zeros,]
res = Reduce(cbind,lapply(1:length(Start_Date), function(i){
thebetas = rep(0,ncol(Riskfactors))
returns_tmp = tmp_matrix[rownames(tmp_matrix) >= Start_Date[i] & rownames(tmp_matrix) <= End_Date[i],i]
RiskF_tmp = Riskfactors[which(rownames(tmp_matrix) >= Start_Date[i] & rownames(tmp_matrix) <= End_Date[i]),,drop=F]
individualVariance = apply(RiskF_tmp,2,sd)
useonly_nonzero_factors = which(individualVariance !=0 )
if (length(useonly_nonzero_factors) > 0){
RiskF_tmp_nonzero = RiskF_tmp[,useonly_nonzero_factors,drop=F]
# REPLACE THESE THREE LINES WITH R'S OLS... to be sure/robust...
thecovariance = cov(returns_tmp, RiskF_tmp_nonzero)
thevariance = var(RiskF_tmp_nonzero)
thebetas[useonly_nonzero_factors] <- solve(thevariance + 0*diag(ncol(thevariance)))%*%matrix(thecovariance,ncol=1) # maybe add a small diagonal for stability?
}
thebetas
}))
rownames(res)<- colnames(Riskfactors)
res
}
Betas_lm <- function(x,y,Dates, EVENT_RETURNS,Risk_Factors){
#data check
if (class(Risk_Factors) != "data.frame") {
dates <- names(Risk_Factors)
Risk_Factors = data.frame(market=Risk_Factors)
rownames(Risk_Factors) <- dates
}
if (sum(sapply(c("Delta" ,"SMB", "HML", "RMW", "CMA", "RF"), function(i) !(i %in% colnames(Risk_Factors))))!=0)
stop("Beta_lm function has wrong risk factors")
if (sum(sapply(colnames(Risk_Factors), function(i) !(i %in% c("Delta" ,"SMB", "HML", "RMW", "CMA", "RF"))))!=0)
stop("Beta_lm function has wrong risk factors")
#create the right formula for lm
RFfield <- which(colnames(Risk_Factors) == "RF")
NotRFfield <- colnames(Risk_Factors)[setdiff(1:ncol(Risk_Factors),RFfield)]
if(length(RFfield) == 1) {
form <- as.formula(paste("(rj - RF) ~ ",paste(NotRFfield,collapse = "+")))
} else {
form <- as.formula(paste("rj ~ ",paste(NotRFfield,collapse = "+")))
}
#Step 1: create the return data
Start_Date <- Dates[which(rownames(Dates) == x),]
End_Date <- Dates[which(rownames(Dates) == y),]
non_zeros <- which(apply(Risk_Factors!=0,1,sum)!=0)
Risk_Factors <- Risk_Factors[non_zeros,,drop=F]
tmp_matrix <- EVENT_RETURNS[non_zeros,]
#Step 2: estimation of betas uses the following function
betafunc <- function(i){
stats = rep(0,length(NotRFfield)+3)
rj = tmp_matrix[rownames(tmp_matrix) >= Start_Date[i] & rownames(tmp_matrix) <= End_Date[i],i]
data = Risk_Factors[rownames(Risk_Factors) >= Start_Date[i] & rownames(Risk_Factors) <= End_Date[i],,drop=F]
data$rj = rj
model = fastLm(form,data=data)
coeff = summary(model)$coefficients[,1]
stats[1:length(coeff)] = coeff #intercept, betas
stats[length(coeff)+1] = ifelse(length(!is.na(model$residuals)) > 2, sd(model$residual[!is.na(model$residuals)]), 0)
stats[length(stats)] = summary(model)$r.squared #R^2
return(stats)
}
#Step 3: apply the function and format the data nicely
res = Reduce(cbind,lapply(1:length(Start_Date), betafunc))
rownames(res)<- c("alpha",NotRFfield,"IVOL","Rsq")
res
}
VOL_lm <- function(x,y,Dates, EVENT_RETURNS){
Start_Date <- Dates[which(rownames(Dates) == x),]
End_Date <- Dates[which(rownames(Dates) == y),]
volfunc <- function(i) {
tmp = EVENT_RETURNS[rownames(EVENT_RETURNS) >= Start_Date[i] & rownames(EVENT_RETURNS) <= End_Date[i],i]
ifelse(sum(tmp!=0) > 3, sd(tmp[tmp!=0]),0)
}
sapply(1:length(Start_Date), function(i) volfunc(i))
}
RET_lm <- function(x,y,Dates, EVENT_RETURNS){
Start_Date <- Dates[which(rownames(Dates) == x),]
End_Date <- Dates[which(rownames(Dates) == y),]
sapply(1:length(Start_Date), function(i) sum(EVENT_RETURNS[rownames(EVENT_RETURNS) >= Start_Date[i] & rownames(EVENT_RETURNS) <= End_Date[i],i]))
}
create_dates <- function(Event_Date) {
Trading.Day = Event_Date + trading_day_after_announce
One.Day.Before <- Event_Date - 1
Five.Day.Before <- Event_Date - 5
One.Month.Before <- One.Day.Before - 30
Three.Month.Before <- One.Day.Before - 90
Six.Month.Before <- One.Day.Before - 180
One.Year.Before <- One.Day.Before - 360
Two.Years.Before <- One.Day.Before - 2*360
One.Month.After <- 30 + Trading.Day
Three.Month.After <- 90 + Trading.Day
Six.Month.After <- 180 + Trading.Day
One.Year.After <- 365 + Trading.Day
Two.Years.After <- 2*365 + Trading.Day
Three.Years.After <- 3*365 + Trading.Day
Four.Years.After <- 4*365 + Trading.Day
Dates <- data.frame(Event_Date,One.Day.Before,Five.Day.Before,One.Month.Before,Three.Month.Before,Six.Month.Before,One.Year.Before,Two.Years.Before,Trading.Day,One.Month.After,Three.Month.After,Six.Month.After,One.Year.After,Two.Years.After,Three.Years.After,Four.Years.After)
colnames(Dates) <- c("Event.Date","One.Day.Before","Five.Day.Before","One.Month.Before","Three.Month.Before","Six.Month.Before","One.Year.Before","Two.Years.Before","Trading.Day","One.Month.After","Three.Month.After","Six.Month.After","One.Year.After","Two.Years.After","Three.Years.After","Four.Years.After")
Dates <- t(Dates)
return(Dates)
}
create_dates_month <- function(Event_Date, allmonths) {
last_available = tail(allmonths,1)
last_available = paste(format(AddMonths(as.Date(last_available),1),"%Y-%m"), "01",sep="-") # Just make it a date in the future
allmonths = sort(allmonths) # just in case
allmonths = str_sub(allmonths, start=1,end=7)
tmp = match(str_sub(Event_Date, start = 1, end = 7), str_sub(allmonths, start=1,end=7))
Trading.Day = ifelse(!is.na(tmp), paste(allmonths[pmin(tmp+1, length(allmonths))], "01",sep="-"), last_available)
One.Month.After = ifelse(!is.na(tmp), paste(allmonths[pmin(tmp+1+1, length(allmonths))], "01",sep="-"), last_available)
Three.Month.After = ifelse(!is.na(tmp), paste(allmonths[pmin(tmp+1+3, length(allmonths))], "01",sep="-"), last_available)
Six.Month.After = ifelse(!is.na(tmp), paste(allmonths[pmin(tmp+1+6, length(allmonths))], "01",sep="-"), last_available)
One.Year.After = ifelse(!is.na(tmp), paste(allmonths[pmin(tmp+1+12, length(allmonths))], "01",sep="-"), last_available)
Two.Years.After = ifelse(!is.na(tmp), paste(allmonths[pmin(tmp+1+24, length(allmonths))], "01",sep="-"), last_available)
Three.Years.After = ifelse(!is.na(tmp), paste(allmonths[pmin(tmp+1+36, length(allmonths))], "01",sep="-"), last_available)
Four.Years.After = ifelse(!is.na(tmp), paste(allmonths[pmin(tmp+1+48, length(allmonths))], "01",sep="-"), last_available)
Dates <- data.frame(Trading.Day,One.Month.After,Three.Month.After,Six.Month.After,One.Year.After,Two.Years.After,Three.Years.After,Four.Years.After)
colnames(Dates) <- c("Trading.Day","One.Month.After","Three.Month.After","Six.Month.After","One.Year.After","Two.Years.After","Three.Years.After","Four.Years.After")
Dates <- t(Dates)
return(Dates)
}
is.crisis <- function(thedate, crisis_years, slack_pre=0, slack_post = 0)
sum(sapply(1:length(crisis_years), function(i) thedate >= AddMonths(crisis_years[[i]][1], - slack_pre*12) & thedate <= AddMonths(crisis_years[[i]][2], slack_post*12)))!=0
plot_crisis_dates <- function(all_ret_values, monthly = 1){
if (class(all_ret_values) == "matrix" | class(all_ret_values) == "data.frame"){
thenames = rownames(all_ret_values)
} else {
thenames = names(all_ret_values)
}
abline(h=0)
for (i in 1:length(BEAR_YEARS)){
if (!monthly){
abline(v=which.min(abs(as.Date(thenames) - BEAR_YEARS[[i]][1])),lwd=3, col= "black", lty="solid")