R MICE 插补
R MICE Imputation
data=data.frame("student"=c(1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5),
"time"=c(1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4),
"v1"=c(16,12,14,12,17,16,12,12,13,12,16,16,10,10,14,17,17,12,10,11),
"v2"=c(1,1,3,2,2,2,3,1,2,1,2,1,3,1,1,2,3,3,1,2),
"v3"=c(4,1,4,4,2,2,2,2,1,3,2,3,1,2,2,1,4,1,1,4),
"v4"=c(NA,27,NA,42,40,48,45,25,29,NA,NA,27,NA,NA,NA,NA,NA,NA,44,39),
"v5"=c(NA,1,NA,NA,1,3,3,2,NA,NA,NA,1,NA,NA,NA,NA,3,2,4,1),
"v6"=c(NA,0,1,NA,1,NA,1,NA,0,NA,1,1,NA,NA,NA,NA,0,0,NA,0),
"v7"=c(0,1,1,NA,0,1,1,0,1,0,NA,0,NA,NA,NA,NA,0,1,NA,1),
"v8"=c(1,NA,0,1,0,0,NA,1,1,NA,0,0,NA,NA,NA,NA,1,0,NA,1))
这是我的样本数据，我想用它来：
一个。对于时间 = 1，使用 v1-v3 使用 MICE 估算 v4-v8（v4 是连续的，v5 是分类的，v6-v8 是二进制的）
乙。在为 time = 1 估算值后，我想用前一个值填充 NA 值。因此，如果时间 1-4 的变量是：NA,NA,0,1 并且时间 1 的估算值为 1，则它可能是：1-1-0-1
我尝试过：
dataNEW <- mice(data[,data$time == 1],m=5,maxit=50,meth='pmm',seed=500)
A. For time = 1 use v1-v3 to impute v4-v8 using MICE (v4 is continuous, v5 is categorical, v6-v8 is binary)
首先，变量 v5 - v6 必须转换为因子：
data$v5 <- factor(data$v5)
data$v6 <- factor(data$v6)
data$v7 <- factor(data$v7)
data$v8 <- factor(data$v8)
创建一个预测矩阵来告诉老鼠只使用 v1-v3 来预测 v4-v8：
Pred_Matrix <- 1 - diag(ncol(data))
Pred_Matrix[,c(1:2, 6:10)] <- 0
仅使用 1 个插补（默认为 5）进行插补，因为您想要的只是插补值；您没有做任何其他事情，例如汇集建模结果。
impA <- mice(subset(data, subset = time==1), pred = Pred_Matrix, m = 1)
可以使用 complete 函数（来自 mice 包，而不是 tidyr）提取估算数据。
B. After imputed values are imputed for time = 1, I want to fill NA
values that follow with the previous value. So if the variable for
time 1-4 is: NA,NA,0,1 and the imputed value at time 1 is 1, then it
could be: 1-1-0-1
library(dplyr)
library(tidyr) # Needed for the fill function
mice::complete(impA) %>%
rbind(subset(data, subset=time!=1)) %>%
arrange(student, time) %>%
group_by(student) %>%
fill(v4:v8)
# A tibble: 20 x 10
# Groups: student [5]
student time v1 v2 v3 v4 v5 v6 v7 v8
<dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <fct> <fct> <fct> <fct>
1 1 1 16 1 4 40 2 1 0 1
2 1 2 12 1 1 27 1 0 1 1
3 1 3 14 3 4 27 1 1 1 0
4 1 4 12 2 4 42 1 1 1 1
5 2 1 17 2 2 40 1 1 0 0
6 2 2 16 2 2 48 3 1 1 0
7 2 3 12 3 2 45 3 1 1 0
8 2 4 12 1 2 25 2 1 0 1
9 3 1 13 2 1 29 1 0 1 1
10 3 2 12 1 3 29 1 0 0 1
11 3 3 16 2 2 29 1 1 0 0
12 3 4 16 1 3 27 1 1 0 0
13 4 1 10 3 1 40 1 0 0 0
14 4 2 10 1 2 40 1 0 0 0
15 4 3 14 1 2 40 1 0 0 0
16 4 4 17 2 1 40 1 0 0 0
17 5 1 17 3 4 40 3 0 0 1
18 5 2 12 3 1 40 2 0 1 0
19 5 3 10 1 1 44 4 0 1 0
20 5 4 11 2 4 39 1 0 1 1
数据
注意，我必须将 v5 的第一个值更改为 2，否则 polyreg 插补失败（time=1 只有两个类别）。
data=data.frame("student"=c(1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5),
"time"=c(1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4),
"v1"=c(16,12,14,12,17,16,12,12,13,12,16,16,10,10,14,17,17,12,10,11),
"v2"=c(1,1,3,2,2,2,3,1,2,1,2,1,3,1,1,2,3,3,1,2),
"v3"=c(4,1,4,4,2,2,2,2,1,3,2,3,1,2,2,1,4,1,1,4),
"v4"=c(NA,27,NA,42,40,48,45,25,29,NA,NA,27,NA,NA,NA,NA,NA,NA,44,39),
"v5"=c(2,1,NA,NA,1,3,3,2,NA,NA,NA,1,NA,NA,NA,NA,3,2,4,1),
"v6"=c(NA,0,1,NA,1,NA,1,NA,0,NA,1,1,NA,NA,NA,NA,0,0,NA,0),
"v7"=c(0,1,1,NA,0,1,1,0,1,0,NA,0,NA,NA,NA,NA,0,1,NA,1),
"v8"=c(1,NA,0,1,0,0,NA,1,1,NA,0,0,NA,NA,NA,NA,1,0,NA,1))
data=data.frame("student"=c(1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5),
"time"=c(1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4),
"v1"=c(16,12,14,12,17,16,12,12,13,12,16,16,10,10,14,17,17,12,10,11),
"v2"=c(1,1,3,2,2,2,3,1,2,1,2,1,3,1,1,2,3,3,1,2),
"v3"=c(4,1,4,4,2,2,2,2,1,3,2,3,1,2,2,1,4,1,1,4),
"v4"=c(NA,27,NA,42,40,48,45,25,29,NA,NA,27,NA,NA,NA,NA,NA,NA,44,39),
"v5"=c(NA,1,NA,NA,1,3,3,2,NA,NA,NA,1,NA,NA,NA,NA,3,2,4,1),
"v6"=c(NA,0,1,NA,1,NA,1,NA,0,NA,1,1,NA,NA,NA,NA,0,0,NA,0),
"v7"=c(0,1,1,NA,0,1,1,0,1,0,NA,0,NA,NA,NA,NA,0,1,NA,1),
"v8"=c(1,NA,0,1,0,0,NA,1,1,NA,0,0,NA,NA,NA,NA,1,0,NA,1))
这是我的样本数据，我想用它来：
一个。对于时间 = 1，使用 v1-v3 使用 MICE 估算 v4-v8（v4 是连续的，v5 是分类的，v6-v8 是二进制的）
乙。在为 time = 1 估算值后，我想用前一个值填充 NA 值。因此，如果时间 1-4 的变量是：NA,NA,0,1 并且时间 1 的估算值为 1，则它可能是：1-1-0-1
我尝试过：
dataNEW <- mice(data[,data$time == 1],m=5,maxit=50,meth='pmm',seed=500)
A. For time = 1 use v1-v3 to impute v4-v8 using MICE (v4 is continuous, v5 is categorical, v6-v8 is binary)
首先，变量 v5 - v6 必须转换为因子：
data$v5 <- factor(data$v5)
data$v6 <- factor(data$v6)
data$v7 <- factor(data$v7)
data$v8 <- factor(data$v8)
创建一个预测矩阵来告诉老鼠只使用 v1-v3 来预测 v4-v8：
Pred_Matrix <- 1 - diag(ncol(data))
Pred_Matrix[,c(1:2, 6:10)] <- 0
仅使用 1 个插补（默认为 5）进行插补，因为您想要的只是插补值；您没有做任何其他事情，例如汇集建模结果。
impA <- mice(subset(data, subset = time==1), pred = Pred_Matrix, m = 1)
可以使用 complete 函数（来自 mice 包，而不是 tidyr）提取估算数据。
B. After imputed values are imputed for time = 1, I want to fill NA values that follow with the previous value. So if the variable for time 1-4 is: NA,NA,0,1 and the imputed value at time 1 is 1, then it could be: 1-1-0-1
library(dplyr)
library(tidyr) # Needed for the fill function
mice::complete(impA) %>%
rbind(subset(data, subset=time!=1)) %>%
arrange(student, time) %>%
group_by(student) %>%
fill(v4:v8)
# A tibble: 20 x 10
# Groups: student [5]
student time v1 v2 v3 v4 v5 v6 v7 v8
<dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <fct> <fct> <fct> <fct>
1 1 1 16 1 4 40 2 1 0 1
2 1 2 12 1 1 27 1 0 1 1
3 1 3 14 3 4 27 1 1 1 0
4 1 4 12 2 4 42 1 1 1 1
5 2 1 17 2 2 40 1 1 0 0
6 2 2 16 2 2 48 3 1 1 0
7 2 3 12 3 2 45 3 1 1 0
8 2 4 12 1 2 25 2 1 0 1
9 3 1 13 2 1 29 1 0 1 1
10 3 2 12 1 3 29 1 0 0 1
11 3 3 16 2 2 29 1 1 0 0
12 3 4 16 1 3 27 1 1 0 0
13 4 1 10 3 1 40 1 0 0 0
14 4 2 10 1 2 40 1 0 0 0
15 4 3 14 1 2 40 1 0 0 0
16 4 4 17 2 1 40 1 0 0 0
17 5 1 17 3 4 40 3 0 0 1
18 5 2 12 3 1 40 2 0 1 0
19 5 3 10 1 1 44 4 0 1 0
20 5 4 11 2 4 39 1 0 1 1
数据
注意，我必须将 v5 的第一个值更改为 2，否则 polyreg 插补失败（time=1 只有两个类别）。
data=data.frame("student"=c(1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5),
"time"=c(1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4),
"v1"=c(16,12,14,12,17,16,12,12,13,12,16,16,10,10,14,17,17,12,10,11),
"v2"=c(1,1,3,2,2,2,3,1,2,1,2,1,3,1,1,2,3,3,1,2),
"v3"=c(4,1,4,4,2,2,2,2,1,3,2,3,1,2,2,1,4,1,1,4),
"v4"=c(NA,27,NA,42,40,48,45,25,29,NA,NA,27,NA,NA,NA,NA,NA,NA,44,39),
"v5"=c(2,1,NA,NA,1,3,3,2,NA,NA,NA,1,NA,NA,NA,NA,3,2,4,1),
"v6"=c(NA,0,1,NA,1,NA,1,NA,0,NA,1,1,NA,NA,NA,NA,0,0,NA,0),
"v7"=c(0,1,1,NA,0,1,1,0,1,0,NA,0,NA,NA,NA,NA,0,1,NA,1),
"v8"=c(1,NA,0,1,0,0,NA,1,1,NA,0,0,NA,NA,NA,NA,1,0,NA,1))