-
Notifications
You must be signed in to change notification settings - Fork 1
/
make_jobs_data.r
55 lines (47 loc) · 2.94 KB
/
make_jobs_data.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
library(tidyverse)
job_org = read.csv('jobs_data/behanghel_processed.csv')
total_reps = 64
for(num_rep in 0:total_reps-1){
job = job_org
# make ipw weights
job = job %>% mutate(ipw = POIDS_PZ_6MOIS / (CVE*mean(job$CVE)+OPP*mean(job$OPP)+CLA*mean(job$CLA)))
# For Z(t) = I[Employed if given t?]
# set outcomes to Y(t) = Z(t) - E[Z(t)]
CVEeff = mean(job$ipw * job$EMPLOI_6MOIS * job$CVE) - mean(job$ipw * job$EMPLOI_6MOIS * job$CLA)
OPPeff = mean(job$ipw * job$EMPLOI_6MOIS * job$OPP) - mean(job$ipw * job$EMPLOI_6MOIS * job$CLA)
CLAbase = mean(job$ipw * job$EMPLOI_6MOIS * job$CLA)
job = job %>% mutate(y=EMPLOI_6MOIS - CVE*CVEeff - OPP*OPPeff - CLAbase)
# make personalization X
Xcat = c('temps', 'rsqstat', 'zus', 'College_education', 'One_to_5_years_of_exp_in_the_job', 'Technician', 'Skilled_clerical_worker', 'Skilled_blue_colar', 'Q1', 'Q2', 'Q3', 'Q4')
for(xc in Xcat){
job[,xc] = factor(job[,xc])
}
Xcts = c()
Xform = paste(Xcat,collapse=" + ")
# make treatment variable
job$t = 1*job$CVE + 2*job$OPP + 3*job$CLA
trnidx = sample(nrow(job),floor(nrow(job)*2/5))
jobtst = job[-trnidx,]
jobtrn = job[trnidx,]
# confound the train data
o = 4
kp = (2/3)/o
jobtrn$z.keep = ((jobtrn$One_to_5_years_of_exp_in_the_job==1)-mean(jobtrn$One_to_5_years_of_exp_in_the_job==1))/sd(jobtrn$One_to_5_years_of_exp_in_the_job==1) + ((jobtrn$Skilled_blue_colar==1)-mean(jobtrn$Skilled_blue_colar==1))/sd(jobtrn$Skilled_blue_colar==1) + ((jobtrn$rsqstat=='RS2')-mean(jobtrn$rsqstat=='RS2'))/sd(jobtrn$rsqstat=='RS2')
jobtrn$p.keep =(
(jobtrn$z.keep<quantile(jobtrn$z.keep,1/3))*(kp/2+jobtrn$CLA*(1-3*kp/2))+
(jobtrn$z.keep>=quantile(jobtrn$z.keep,2/3))*(kp/2+jobtrn$CVE*(1-3*kp/2))+
((jobtrn$z.keep>=quantile(jobtrn$z.keep,1/3)) & (jobtrn$z.keep<quantile(jobtrn$z.keep,2/3)))*(kp/2+jobtrn$OPP*(1-3*kp/2))
)
jobtrn$keep = runif(nrow(jobtrn)) <= jobtrn$p.keep
jobtrn = jobtrn[jobtrn$keep,]
train_data = data.frame(temps=jobtrn$temps, rsqstat=jobtrn$rsqstat, zus=jobtrn$zus, College_education=jobtrn$College_education,
One_to_5_years_of_exp_in_the_job=jobtrn$One_to_5_years_of_exp_in_the_job, Technician=jobtrn$Technician,
Skilled_clerical_worker=jobtrn$Skilled_clerical_worker, Skilled_blue_colar=jobtrn$Skilled_blue_colar,
Q1=jobtrn$Q1, Q2=jobtrn$Q2, Q3=jobtrn$Q3, Q4=jobtrn$Q4, t=jobtrn$t, ipw=jobtrn$ipw, y=jobtrn$y)
write.csv(train_data, sprintf("jobs_data/train_%d.csv", num_rep), row.names=FALSE)
test_data = data.frame(temps=jobtst$temps, rsqstat=jobtst$rsqstat, zus=jobtst$zus, College_education=jobtst$College_education,
One_to_5_years_of_exp_in_the_job=jobtst$One_to_5_years_of_exp_in_the_job, Technician=jobtst$Technician,
Skilled_clerical_worker=jobtst$Skilled_clerical_worker, Skilled_blue_colar=jobtst$Skilled_blue_colar,
Q1=jobtst$Q1, Q2=jobtst$Q2, Q3=jobtst$Q3, Q4=jobtst$Q4, t=jobtst$t, ipw=jobtst$ipw, y=jobtst$y)
write.csv(test_data, sprintf("jobs_data/test_%d.csv", num_rep), row.names=FALSE)
}