forked from NSAPH/GPSmatching
/
generate_synthetic_data.R
executable file
·181 lines (146 loc) · 6.11 KB
/
generate_synthetic_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
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
#' @title
#' Generate synthetic data for the CausalGPS package
#'
#' @description
#' Generates synthetic data set based on different GPS models and covariates.
#'
#' @param sample_size A positive integer number that represents a number of data
#' samples.
#' @param outcome_sd A positive double number that represents standard deviation
#' used to generate the outcome in the synthetic data set.
#' @param gps_spec A numerical integer values ranging from 1 to 7. The
#' complexity and form of the relationship between covariates and treatment
#' variables are determined by the `gps_spec`. Below, you will find a concise
#' definition for each of these values:
#' - *gps_spec: 1*: The treatment is generated using a normal distributionMay 24, 2023
#' (`stats::rnorm`) and a linear function of covariates (cf1 to cf6).
#' - *gps_spec: 2*: The treatment is generated using a Student's
#' t-distribution (`stats::rt`) and a linear function of covariates, but is
#' also truncated to be within a specific range (-5 to 25).
#' - *gps_spec: 3*: The treatment includes a quadratic term for the third
#' covariate.
#' - *gps_spec: 4*: The treatment is calculated using an exponential
#' function within a fraction, creating logistic-like model.
#' - *gps_spec: 5*: The treatment also uses logistic-like model but with
#' different parameters.
#' - *gps_spec: 6*: The treatment is calculated using the natural logarithm
#' of the absolute value of a linear combination of the covariates.
#' - *gps_spec: 7*: The treatment is generated similarly to `gps_spec = 2`,
#' but without truncation.
#' @param cova_spec A numerical value (1 or 2) to modify the covariates. It
#' determines how the covariates in the synthetic data set are transformed.
#' If `cova_spec` equals 2, the function applies non-linear transformation to
#' the covariates, which can add complexity to the relationships between
#' covariates and outcomes in the synthetic data. See the code for more details.
#'
#' @param vectorized_y A Boolean value indicates how Y internally is generated.
#' (Default = `FALSE`). This parameter is introduced for backward compatibility.
#' vectorized_y = `TRUE` performs better.
#'
#'
#' @return
#' \code{synthetic_data}: The function returns a data.frame saved the
#' constructed synthetic data.
#'
#' @export
#'
#' @examples
#'
#' set.seed(298)
#' s_data <- generate_syn_data(sample_size = 100,
#' outcome_sd = 10,
#' gps_spec = 1,
#' cova_spec = 1)
#'
generate_syn_data <- function(sample_size = 1000,
outcome_sd = 10,
gps_spec = 1,
cova_spec = 1,
vectorized_y = FALSE) {
if (sample_size < 0 || !is.numeric(sample_size)) {
stop(paste("'sample_size' should be a positive ineteger number.",
" You entered: ", sample_size))
}
if (outcome_sd < 0){
stop(paste("'outcome_sd' should be a positive double number.",
" You entered:", outcome_sd))
}
if(!gps_spec %in% 1:7){
stop(paste("gps_spec: ", gps_spec, ", is not a valid value."))
}
if(!cova_spec %in% 1:2){
stop(paste("cova_spec: ", cova_spec, ", is not a valid value."))
}
size <- sample_size
#pre-treatment variables (confounders)
cf <- MASS::mvrnorm(n = size,
mu = c(0,0,0,0),
Sigma = matrix(c(1,0,0,0,0,1,0,0,0,0,1,0,0,0,0,1),
ncol=4))
cf5 <- sample(c((-2):2), size, replace = TRUE)
cf6 <- stats::runif(size, min=-3, max=3)
tmp_treat <- (- 0.8 + 0.1 * cf[, 1]
+ 0.1 * cf[, 2]
- 0.1 * cf[, 3]
+ 0.2 * cf[, 4]
+ 0.1 * cf5
+ 0.1 * cf6)
if (gps_spec == 1) {
treat <- (tmp_treat * 9 + 17 + stats::rnorm(size,sd=5))
} else if (gps_spec == 2) {
treat <- (tmp_treat * 15 + 22 + stats::rt(size,2))
treat[which(treat < (-5))] <- (-5)
treat[which(treat > (25))] <- (25)
} else if (gps_spec == 3) {
treat <- (tmp_treat * 9 + 1.5 * cf[ , 3] ^ 2 +
stats::rnorm(size, mean = 0, 5) + 15)
} else if (gps_spec == 4) {
treat <- (49 * exp(tmp_treat)
/ (1 + exp(tmp_treat)) - 6 + stats::rnorm(size, sd=5))
} else if (gps_spec == 5) {
treat <- (42 / (1 + exp(tmp_treat)) - 18 + stats::rnorm(size,sd=5))
} else if (gps_spec == 6) {
treat <- (log(abs(tmp_treat)) * 7 + 13 + stats::rnorm(size,sd=4))
} else if (gps_spec == 7) {
treat <- (tmp_treat * 15 + 22 + stats::rt(size,2))
} else {
stop(paste("gps_spec: ", gps_spec, ", is not a valid value."))
}
if (vectorized_y){
# Pre-calculate some parts of the formula
cf_sum <- rowSums(cf * c(0.2, 0.2, 0.3, -0.1))
cf_part <- 0.1 + 0.1 * cf[,4] + 0.1 * cf5 + 0.1 * cf[,3] ^ 2
# Pre-generate a vector of random numbers
rand_norm <- stats::rnorm(size, mean = 0, sd = outcome_sd)
# Calculate Y
Y <- -((1 + cf_sum * 10 - 2 * cf5 - 2 * cf6 + (treat - 20) * cf_part
- 0.13 ^ 2 * (treat - 20) ^ 2)) +
rand_norm
} else {
# TODO: This part is kept for backward compatibility.
Y <- as.numeric()
# non-vectorized Y
for (i in 1:size) {
Y[i] <- ((-(1 + (sum(c(0.2, 0.2, 0.3, -0.1) * cf[i, ]))
* 10 - 2 * cf5[i] - 2 * cf6[i] + (treat[i] - 20)
* (0.1 + 0.1 * cf[i,4] + 0.1 * cf5[i]
+ 0.1 * cf[i,3] ^ 2 - 0.13 ^ 2 * (treat[i] - 20) ^ 2)))
+ stats::rnorm(1, mean = 0, sd = outcome_sd))
}
}
if (cova_spec == 1) {
cf <- cf
} else if (cova_spec == 2) {
cf[,1] <- exp(cf[ ,1] / 2)
cf[,2] <- (cf[ ,2] / (1 + exp(cf[ ,1]))) + 10
cf[,3] <- (cf[ ,1] * cf[ ,3]/25 + 0.6) ^ 3
cf[,4] <- (cf[ ,2] + cf[ ,4] + 20) ^ 2
} else {
stop(paste("cova_spec: ", cova_spec, ", is not a valid value."))
}
w <- treat
simulated_data <- data.frame(cbind(Y, w, cf, cf5, cf6))
colnames(simulated_data)[3:8]<-c("cf1", "cf2", "cf3", "cf4", "cf5", "cf6")
simulated_data$id <- seq_along(1:nrow(simulated_data))
return(simulated_data)
}