-
Notifications
You must be signed in to change notification settings - Fork 0
/
B2. modelTempDropout_BART_validationLeaveOneWaveOut.R
165 lines (153 loc) · 6.79 KB
/
B2. modelTempDropout_BART_validationLeaveOneWaveOut.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
################################################################################
################################################################################
## BART FOR PREDICTING PROBABILITIES OF TEMPORARY DROPOUT
## Level: school context
## Validation on level of waves
## 05.11.2019
## Sabine Zinn
################################################################################
################################################################################
rm(list=ls())
set.seed(543)
################################################################################
# I. LOAD LIBRARIES AND DATA
################################################################################
library(BART, lib.loc="Z:\\Projects\\p000139_Methoden_Survey_Psych\\BART_Project")
library(foreign)
library(mice)
setwd("Z:\\Projects\\p000139_Methoden_Survey_Psych\\BART_Project\\1. Prepare data")
load("data.RData")
N <- length(unique(datl$ID_t)) # N=4559
time <- rep(1:9, N)
datl <- cbind(datl, time)
Dat <- datl[datl$time %in% 2:5,]
time <- rep(1:4, N)
Dat$time <- time
id.risk <- Dat[Dat$wave==2 & Dat$indNV==0 ,1] # only take those students into the risk set who did not perm. dropout already in wave 2
Dat <- Dat[Dat$ID_t %in% id.risk,]
N <- length(unique(Dat$ID_t)) # N=4266
md.pattern(Dat, plot=FALSE) # very few missing values in books & mig
# Add total number of participation events as (potential) explanatory variable
Dat$numPart <- NA
addNumberParticiaption <- function(id){
st <- Dat$status[Dat$ID_t %in% id]
st <- cumsum(ifelse(st == 0,1,0)) # only participation events count
Dat$numPart[Dat$ID_t %in% id] <<- rep(st[4], 4) # consider only four waves
return(NULL)
}
nn <- sapply(unique(Dat$ID_t), addNumberParticiaption)
################################################################################
# II. DEFINE MODEL MATRICES AND SETTINGS
################################################################################
# Matrix of covariates as training data
# 1. Define training and test sets (on the level of waves)
D.train <- Dat[Dat$time %in% 1:3,]
x.train <- D.train[!duplicated(D.train$ID_t), c(2:84)]
colnames(x.train) <- colnames(Dat[,c(2:84)])
id.train <- unique(D.train$ID_t)
# Define matrix for times and events, i.e. temporary dropout
L <- 3
times <- NULL
delta <- NULL
for(i in 1:length(id.train)){
#i <- 1
#cat("It: ",i,"\n")
id <- id.train[i]
mm <- D.train[D.train$ID_t %in% id,]
if(1 %in% mm$indNV) { # drop those already dopped out permanently -> this produces an unbalanced panel
fe <- which(mm$indNV %in% 1)[1]
mm$indNV[fe:nrow(mm)] <- 1
wi <- which(mm$indNV %in% 1)
mm <- mm[-wi,]
}
if(nrow(mm)>0){ # count events
time.i <- 1:nrow(mm)
delta.i <- rep(0, nrow(mm))
if(1 %in% mm$status){ # there is an event / are events "temp. dropout"
td.i <- which(mm$status %in% 1)
delta.i[td.i] <- 1
}
if(length(time.i)<L)
time.i <- c(time.i, rep(0, L-length(time.i)))
if(length(delta.i)<L)
delta.i <- c(delta.i, rep(0, L-length(delta.i)))
times <- rbind(times, c(id, time.i))
delta <- rbind(delta, c(id, delta.i))
rm(time.i); rm(delta.i)
}
}
apply(delta[,-1],2,sum)
xx <- apply(times[,-1],1,paste, collapse="-")
xy <- apply(delta[,-1],1,paste, collapse="-")
table(xx,xy) # empirical distribution of events is okay
sum(apply(delta[,-1],2,sum)) # this is the number of temporary dropouts in school in the traning data set (N=589)
# Define init structure for BART
pre <- recur.pre.bart(times=times[,-1], delta=delta[,-1], x.train= as.matrix(x.train))
################################################################################
# III. MODEL ESTIMATION
################################################################################
post <- recur.bart(y.train=pre$y.train, pre$tx.train, x.test=pre$tx.test,
# ntree = 300, usequants=TRUE, nskip=200, sparse=TRUE,
# ndpost=1500, keepevery=200)
ntree = 300, usequants=TRUE, nskip=500, sparse=TRUE,
ndpost= 5000, keepevery=500)
save.image("Z:\\Projects\\p000139_Methoden_Survey_Psych\\BART_Project\\Results\\partModel_est_validWave.Rdata")
################################################################################
# V. VALIDATION on the level of wave 4
# (in paper this refers to wave 5, here counting starts at wave 0)
################################################################################
# Predict values for all waves and then take only the last one for validation
ids <- unique(Dat$ID_t)
# Define matrix for times and events, i.e. temporary dropout
L <- 4
times <- NULL
delta <- NULL
for(i in 1:length(ids)){
#i <- 1
#cat("It: ",i,"\n")
id <- ids[i]
mm <- Dat[Dat$ID_t %in% id,]
if(1 %in% mm$indNV) { # drop those already dopped out permanently -> this produces an unbalanced panel
fe <- which(mm$indNV %in% 1)[1]
mm$indNV[fe:nrow(mm)] <- 1
wi <- which(mm$indNV %in% 1)
mm <- mm[-wi,]
}
if(nrow(mm)>0){ # count events
time.i <- 1:nrow(mm)
delta.i <- rep(0, nrow(mm))
if(1 %in% mm$status){ # there is an event / are events "temp. dropout"
td.i <- which(mm$status %in% 1)
delta.i[td.i] <- 1
}
if(length(time.i)<L)
time.i <- c(time.i, rep(0, L-length(time.i)))
if(length(delta.i)<L)
delta.i <- c(delta.i, rep(0, L-length(delta.i)))
times <- rbind(times, c(id, time.i))
delta <- rbind(delta, c(id, delta.i))
rm(time.i); rm(delta.i)
}
}
apply(delta[,-1],2,sum)
xx <- apply(times[,-1],1,paste, collapse="-")
xy <- apply(delta[,-1],1,paste, collapse="-")
table(xx,xy) # empirical distribution of events is okay
sum(apply(delta[,-1],2,sum)) # this is the number of temporary dropouts in school in the traning data set (N=589)
# Define init structure for BART
x.test <- Dat[!duplicated(Dat$ID_t), c(2:84,86)]
pre <- recur.pre.bart(times=times[,-1], delta=delta[,-1], x.train= as.matrix(x.test))
preX <- pre$tx.train[,post$rm.const]
preXP <- preX[,!(colnames(preX) %in% "status")]
predV <- predict(post,preXP) # on the observed
head(predV$prob.test[,1:20])
# Derive posterior draws of prob. of perm. dropout for test set, for each wave 4 only
pi.4 <- predV$prob.test[,preX[,"t"]==4]
# Get accuracy for each wave
preX <- as.data.frame(preX)
preX$tempD <- ifelse(preX$status %in% 1,1,0)
prop.table(table(preX[,"tempD"], preX[,"t"]),2) # empirical (observed) distribution of permanent dropout events
table(preX[,"t"]) # risk set decreases from wave to wave
dr4 <- mean(apply(pi.4,1,mean)) # Wave 4, predicted probability to dropout temporarily in wave 4
bart_predict_wave4 <- ifelse(apply(pi.4,2,mean)>0.5,1,0) # accuracy
print(mean(bart_predict_wave4==preX[,"tempD"][preX[,"t"]==4])) # 0.97