Subset multiple times in a data frame
I want a 100 times subset of a dataframe made up of 20 variables (continuous and categorical) in two parts that are 70% and 30%. But I can do it with aperture dataset as an example as well.
data(iris)
test.rows <- sample(1:nrow(iris), 105)
iris.70 <- iris[test.rows, ]
iris.30 <- iris[-test.rows, ]
This gives me the dataframes I need. But how can I do this 100 times and save the results somewhere so I can use them later?
I have tried
output <- list()
for(i in 1:surveyed100){
output[[i]] <- test.rows <- sample(1:nrow(surveyed100), 246)
}
But this tells me: a numeric expression has 5 elements: only the first one is used.
I would appreciate your help.
source to share
First, create 100 samples:
samples <- list()
for(i in 1:100){
samples[[i]] <- sample(1:nrow(surveyed100), 246)
}
Then use lapply()
to store all 100 subsets in the list:
output <- lapply(samples, function(x) list(surveyed100[x,], surveyed100[-x,]))
Usage example iris
:
samples <- list()
for(i in 1:100){
samples[[i]] <- sample(1:nrow(iris), 105)
}
head(samples)
[[1]]
[1] 66 106 39 50 33 123 68 62 65 125 30 25 60 70 49 98 140 44 141 94 18 59 117 32 63 133 16 139 97 145 105 78 112 95
[35] 128 36 37 64 10 124 40 111 17 29 51 89 99 4 135 103 101 19 115 74 73 91 11 67 84 88 1 114 138 21 77 24 69 13
[69] 53 58 110 150 9 31 144 54 129 34 35 52 142 14 113 127 27 20 87 134 118 15 72 92 75 8 104 96 136 143 2 41 109 90
[103] 146 26 6
[[2]]
[1] 78 84 89 75 63 81 119 51 127 20 66 106 140 65 116 72 147 141 61 113 130 136 109 49 57 149 90 56 8 46 82 55 38 4
[35] 70 94 100 117 95 29 45 13 128 11 83 80 35 41 121 73 39 67 19 98 108 103 42 2 44 132 114 137 118 12 125 24 77 53
[69] 28 150 92 5 43 112 60 122 15 30 104 102 120 76 47 85 40 79 33 143 48 139 148 124 36 16 138 101 115 107 134 126 74 6
[103] 52 50 10
[[3]]
[1] 23 67 54 131 84 146 25 7 41 101 138 49 28 95 15 5 57 69 126 60 12 92 35 89 50 1 13 77 140 116 136 17 144 64
[35] 32 139 76 102 61 130 2 44 75 100 81 31 34 46 72 33 18 79 24 133 124 62 9 88 8 66 74 125 51 127 123 52 90 39
[69] 120 42 16 83 40 137 47 58 82 135 96 20 119 91 36 48 132 55 93 106 107 109 113 53 19 141 105 128 78 143 29 4 45 37
[103] 73 94 87
[[4]]
[1] 125 41 37 80 136 50 91 89 44 117 132 82 78 128 146 49 61 105 145 83 111 126 100 94 7 102 112 17 120 60 36 104 123 65
[35] 48 34 45 73 25 46 110 74 66 137 107 101 106 24 97 18 119 72 33 134 87 35 121 14 88 9 39 8 64 142 10 148 54 99
[69] 103 95 63 11 133 141 32 96 51 81 140 76 138 127 52 75 55 26 115 19 90 16 21 86 56 22 79 53 31 23 68 13 77 30
[103] 71 116 67
[[5]]
[1] 83 4 85 133 111 55 145 65 81 50 136 64 13 27 5 117 33 69 40 127 80 61 53 125 77 36 124 140 138 86 7 6 79 29
[35] 21 115 23 74 93 10 132 51 2 41 49 123 94 142 120 48 19 89 28 91 14 118 43 103 87 58 149 20 56 113 82 62 104 44
[69] 72 47 119 35 143 116 128 26 75 88 9 60 16 130 114 31 1 147 78 73 3 32 70 146 131 102 15 54 141 129 42 101 17 59
[103] 46 134 110
[[6]]
[1] 18 20 53 106 142 125 120 109 119 129 84 146 99 51 43 91 141 89 131 124 95 135 81 42 73 112 128 133 108 27 28 47 32 76
[35] 130 138 70 36 10 90 16 11 137 17 87 5 35 25 123 97 12 115 127 94 34 103 4 54 134 78 68 71 101 126 61 37 33 2
[69] 88 80 144 82 150 3 21 114 58 110 136 22 105 117 79 64 102 49 98 59 132 39 8 149 121 40 29 104 55 77 147 74 50 56
[103] 48 75 23
subsets:
output <- lapply(samples, function(x) list(iris[x,], iris[-x,]))
Output:
head(output[[1]][[1]])
Sepal.Length Sepal.Width Petal.Length Petal.Width Species
66 6.7 3.1 4.4 1.4 versicolor
106 7.6 3.0 6.6 2.1 virginica
39 4.4 3.0 1.3 0.2 setosa
50 5.0 3.3 1.4 0.2 setosa
33 5.2 4.1 1.5 0.1 setosa
123 7.7 2.8 6.7 2.0 virginica
head(output[[1]][[2]])
Sepal.Length Sepal.Width Petal.Length Petal.Width Species
3 4.7 3.2 1.3 0.2 setosa
5 5.0 3.6 1.4 0.2 setosa
7 4.6 3.4 1.4 0.3 setosa
12 4.8 3.4 1.6 0.2 setosa
22 5.1 3.7 1.5 0.4 setosa
23 4.6 3.6 1.0 0.2 setosa
> nrow(output[[1]][[1]])
[1] 105
> nrow(output[[1]][[2]])
[1] 45
source to share
You can create a small function to do this, for example:
foo <- function(dat, train_percent = 0.7) {
n <- seq_len(nrow(dat))
train <- sample(n, floor(train_percent * max(n)))
test <- sample(setdiff(n, train))
list(train = dat[train,], test = dat[test,])
}
Then you can easily apply this function multiple times using replicate
:
replicate(100, foo(iris), simplify = FALSE)
The result list contains 100 items, and each item is itself a list of two items, where the first is the "train" and the second is the "test" dataset.
source to share