-
Notifications
You must be signed in to change notification settings - Fork 1
/
General Data Prep 150304B.R
230 lines (183 loc) · 10.8 KB
/
General Data Prep 150304B.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
#########################################################################
#
# General data Prep
# consilidates most of data prep that doesn't require rsq (aplicable question list)
#
#############################################################################
# setwd("C:/Users/Walter/Documents/GitHub/SciCast_Andy/SciCast_Andy")
#
# First run Get_Data.R.
startGen <- Sys.time()
print("Data Generation started")
tstart <- as.POSIXct("2013-11-25 00:00:00 EST")
base <- tstart-28*24*60*60
tstop <- Sys.time()
days <- seq(1,ceiling(as.double(tstop - tstart)),1)
source("Incentive Selection 150312.R")
#
# Removing admin accounts and activity (Data_cleaning)
# Match to Steve's list!
pip <- pr$user_id; pus <- as.character(pr$username); cap <- as.POSIXct(pr$created_at); grps <- pr$groups; rip <-pr$referral_id
adu <- c("amsiegel","BAE11","brnlsl","brobins","cedarskye","christinafreyman","ctwardy",
"daggre_admin","dquere","gbs_tester","Inkling","jessiejury","jlu_bae","kennyth0",
"klaskey","kmieke","manindune","Naveen Jay","pthomas524","Question_Admin",
"Question Mark","randazzese","RobinHanson","saqibtq","scicast_admin","slin8",
"ssmith","tlevitt","wsun")
ads <- integer()
ads <- c(15,16,70,82,135,249,855) #specificly excluded user_ids from MITRE
######## All are coded as internal users, so why is this part needed? #########
adi <- numeric() #adi in a list of specifically excluded pr_user_names
for (i in 1:length(adu)) {
adi[i] <- pip[pus==adu[i]]
cap[pip==adi[i]] <- NA #cap is pr_created_at with NAs for Pr_user_ids on Adi list
}
for (i in 1:length(ads)) {
cap[pip==ads[i]] <- NA #puts NA for each pr_created_at for each admin user
adi <- c(adi,ads[i])
}
grp <- array(rep("a",length(pip)*20),c(length(pip),20)); igrp <- rep(0,length(pip)) #grp is an array of all groups a user is part of
for (i in 1:length(pip)) {
temp <- as.vector(strsplit(as.character(grps[i]),",")[[1]]) #makes a vector of the list of groups
grp[i,1:length(temp)] <- temp #puts the vector in the grp array
}
#adi <- numeric()
for (i in 1:length(pip)) {
for (g in 1:20) {
if (grp[i,g]=="Admin"|grp[i,g]=="SuperAdmin"|grp[i,g]=="UserAdmin"|grp[i,g]=="BadgesAdmin"|grp[i,g]=="RolesAdmin"|grp[i,g]=="QuestionAdmin") {
cap[i] <- NA #puts NA for each pr_created_at for each admin user
adi <- c(adi,pip[i]) #adds admion users to adi list
}
if (grp[i,g]=="Internal") { # igrp is a list (length(pip)) with internal users flagged with 1s
igrp[i] <- 1
adi <- c(adi,pip[i])
}
}
}
adi <- unique(adi)
good <- complete.cases(cap)
sum(!good) # How many are not good?
cap<-cap[good]; pus<-pus[good]; pip<-pip[good]; grp<-grp[good,]; rip<-rip[good]
rust <- as.character(th$raw_user_selection)
rust[rust=="[\"\\\"Will Not occur by December 31, 2034\\\"\",[0.9545454545454546,1],null]"] <- "[\"\\\"Will Not occur by December 31 2034\\\"\",[0.9545454545454546,1],null]"
rust[rust=="[\"\\\"Will Not occur by December 31, 2034\\\"\", [0.9545454545454546, 1], null]"] <- "[\"\\\"Will Not occur by December 31 2034\\\"\", [0.9545454545454546,1], null]"
rust[rust=="[\"\\\"Less than 4.75%\\\"\",[-0.33333333333333326,-0.33333333333333326],null]"] <- "None"
rust[rust=="[\"\\\"Between 4.75% and 5%\\\"\",[-0.33333333333333326,0.33333333333333326],\"Lower\"]"] <- "None"
rust[rust=="[\"\\\"Less than 4.75%\\\"\", [-0.33333333333333326, -0.33333333333333326], null]"] <- "None"
rust[rust=="[\"Down ~12% or more\",[-0.21428571428571427,0.5],null]"] <- "None"
rust[rust=="[\"Up ~12% or more\",[0.842857142857143,1.2142857142857142],null]"] <- "None"
rst <- rep(0,length(rust))
mdt <- rep(1,length(rust)) ##orinially rst <- med <- rep(0,length(rust))
#for (t in 1:39) {
for (t in 1:length(rust)) {
#if (grepl("None",rust[t])) {
if (thInterface[t]==2) {
mdt[t] = 0 # 1 indicates safe-mode forecast.
tmp1 <- as.double(strsplit(as.vector(nvt[t]),",")[[1]])
rst[t] <- tmp1[cit[t]+1]
} else {
# A later selection on SciCast.org like "Higher" will assume the user wants the forecast halfway between the current market estimate and the top of the bin.
tmp1 <- as.double(strsplit(as.vector(ovt[t]),",")[[1]])
#rust -> ["<str1>",[<num1>,<num2>],"<str2>"] ["<str1>",[<num1>,<num2>],null]
if (grepl("Lower", rust[t])) { rst[t] <- (tmp1[cit[t]+1] + (as.double(strsplit(strsplit(rust[t],',')[[1]][2],'[',fixed=T)[[1]][2])))/2}
##strsplit(rust[t],',')[[1]][2] -> [<num1>
##strsplit([<num1>,']',fixed=T)[[1]][1]) -> <num1> (bin min)
##tmp2 -> x - (x- binmax)/2
if (grepl("Higher", rust[t])) { rst[t] <- (tmp1[cit[t]+1] + (as.double(strsplit(strsplit(rust[t],',')[[1]][3],']',fixed=T)[[1]][1])))/2 }
##strsplit(rust[t],',')[[1]][3] -> <num2>]
##strsplit(<num2>],']',fixed=T)[[1]][1]) -> <num2> (bin max)
##tmp2 -> x + (binmax - x)/2
#if (tmp2=="Higher") { rst[t] <- tmp1[cit[t]+1] + (as.double(strsplit(strsplit(rust[t],',')[[1]][3],']',fixed=T)[[1]][1]) -tmp1[cit[t]+1])/2 }
##strsplit(rust[t],',')[[1]][3] -> <num2>]
##strsplit(<num2>],']',fixed=T)[[1]][1]) -> <num2>
##tmp2 -> tmp1[cit[t]+1] + <num2> -tmp1[cit[t]+1])/2 -> <num2> + tmp1[cit[t]+1])/2
if (grepl("What they are now",rust[t])) { rst[t] <- tmp1[cit[t]+1]}
if (grepl("null",rust[t])) { rst[t] <- mean(as.double(c( strsplit(strsplit(rust[t],",")[[1]][2],"[",fixed=T)[[1]][2], strsplit(strsplit(rust[t],",")[[1]][3],"]",fixed=T)[[1]][1] ))) }
#tmp2 -> mean(<num1. <num2>)
}
}
for (i in 1:length(adi)) {
tat[pit==adi[i]] <- NA
}
good <- complete.cases(tat)
sum(!good) # How many are not good?
tat<-tat[good]; pit<-pit[good]; qit<-qit[good]; nvt<-nvt[good]; ovt<-ovt[good]; ast<-ast[good]; apot<-apot[good]
tit<-tit[good]; cit<-cit[good]; rst<-rst[good]; mdt<-mdt[good]
pic <- cm$user_id
cac <- as.POSIXct(cm$created_at)
qic <- cm$question_id
for (i in 1:length(adi)) {
cac[pic==adi[i]] <- NA
}
good <- complete.cases(cac)
sum(!good) # How many are not good?
cac<-cac[good]; pic<-pic[good]; qic<-qic[good]
lp <- length(pip)
##############
# Setup
##############
#qn$provisional_settled_at is start of comment period and qn$pending_until will be reused for event resolution (not question resolution/settlement)(Analysis_Setup)
drq <- as.double(raq-caq)
hist(drq) # Range of questions' duration
########## Removing forecasts that occurred after resolution was known (Analysis_setup).
startTat <- Sys.time()
print("tat removal started")
for (t in 1:length(tat)) { # for all tardes....
if (tat[t]>raq[qiq==qit[t]]) { # if traded_at > pending_until... #####is pending_until correct vice resolved_at?
tat[t] <- NA # NA subistuted for traded_at
}
}
tat[tat>tstop] <- NA # For all traded_at > than tstop (Sys.time()) substitute Na for traded_at
good <- complete.cases(tat) # remove all trades with trades later than tstop or pending_until
sum(!good)
tat<-tat[good]; tit<-tit[good];pit<-pit[good]; qit<-qit[good]; nvt<-nvt[good]; ovt<-ovt[good]; ast<-ast[good]; apot<-apot[good]
cit<-cit[good]; rst<-rst[good]; mdt<-mdt[good]
duration <- as.double(difftime(Sys.time(),startTat,units="sec")) #reports time to retrieve files
print(c("tat removal complete", duration))
########## generates list of condional assumption questions (asqt) and options (asot), -1 indicates in either indicades non-conditional trade (Analysis_setup)
nowish <- strsplit(as.character(tstop), ' ')[[1]][1] # tstop date only
asq <- aso <- rep("a",length(tat))
for (t in 1:length(tat)) { # for all remaining forecasts...
asq[t] <- strsplit(as.character(ast[t]),':')[[1]][1] # assumptin question as str
aso[t] <- strsplit(as.character(ast[t]),':')[[1]][2] # assumtion option as str
}
asqt <- as.double(asq) # assumptin question as double
asot <- as.double(aso) # assumptin option as double
asqt[is.na(asqt)==T] <- -1 # if no seirialized _assumtions question replace NAs wiht -1
asot[is.na(asot)==T] <- -1 # if no seirialized _assumtions option replace NAs wiht -1 => for non-conditional trades asqt & asot = -1
###########
# For Steve Stratman, DON'T remove stuttered forecasts. (no "de-stuttering")
# Reordering to simplify other operations later (Analysis_Setup).
ord <- order(qit,tat)
tat<-tat[ord]; tit<-tit[ord]; pit<-pit[ord]; qit<-qit[ord]; nvt<-nvt[ord]; ovt<-ovt[ord]; ast<-ast[ord]; apot<-apot[ord]
cit<-cit[ord]; rst<-rst[ord]; mdt<-mdt[ord]; asqt<-asqt[ord]; asot<-asot[ord]
#
# Market Accuracy
# Binary and ordered means continuous; it makes no difference to BS, but it does make a difference on "poco" and "hit".
# ONLY binary for Stratman.
# Find resolved questions (Anaysis_Setup) ##### Not used for input to anything else? ####
'%ni%' <- Negate('%in%')
gpq <- matrix(rep("a",length(qiq)*200),c(length(qiq),200))
vldq <- rep(0,length(qiq)) # initialize
for (q in 1:length(qiq)) {
tmp <- as.vector(strsplit(grq[q],',',fixed=T)[[1]]) # as vector groups associated wiht each question
lv <- length(tmp)
if (lv>0) { gpq[q,1:lv] <- tmp } # If there is a group (and there always is), vector of groups put into gpq
if ("Invalid Questions"%ni%tmp) { vldq[q] <- 1 } # If "Invlid Questions" in list of groups, valid questions = -1
}
# How many are invalid?
length(vldq[vldq==0])
svt <- rvqt <- rvqat<- array(rep(0,length(tat)*40),c(length(tat),40))
roqt <- roqat <-rep(-1,length(tat))
########
# Cleaning up question groups - creating vectors of groups for each user (Data_Cleaning)
#wtg <- rep(0,length(qiq))
gq <- array(numeric(), c(length(qiq),200)) #### WHy not a matrix ####
for (j in 1:length(qiq)) { # For all questions
if(grq[j]!="") { # If there are any groups.....
temp <- levels(factor(strsplit(as.character(grq[j]),",")[[1]])) # generate a vector of groups for each user #### Same as above ###
# wtg[j]<- 1/length(temp) # weight is inversly proportinal to number fo groups
gq[j,1:(length(temp))] <- temp # putting the vector into array gq #### WHy is gq different from gpq
}
}
duration <- as.double(difftime(Sys.time(),startGen,units="sec")) #reports time to retrieve files
print (c("Data Generation Complete", duration))