Split data frame by class regarding to OID - r

Question

I try to split dataframe by 50% by class. However, I do not want to split fields with the same OID (object identifier). I would like the fields with the same OID to be in the same set.
#Data frame:
"b1""b2""b3""CLASS" "OID"
110 134 119 "tree" 1
112 133 118 "tree" 1
105 125 110 "tree" 2
112 132 117 "tree" 2
109 125 115 "meadow" 6
93 110 101 "meadow" 6
86 106 95 "meadow" 7
105 136 116 "meadow" 7
102 128 111 "meadow" 8
108 129 115 "meadow" 8
113 134 119 "meadow" 8
Expected data:
#Expected:
"b1""b2""b3""CLASS" "OID"
110 134 119 "tree" 1
112 133 118 "tree" 1
109 125 115 "meadow" 6
93 110 101 "meadow" 6
86 106 95 "meadow" 7
105 136 116 "meadow" 7

Answer 1

This selects the top half of rows in each group, plus any rows which have the same OID as the rows in that top half.
library(dplyr)
df %>%
group_by(CLASS) %>%
filter(OID %in% head(OID, n() %/% 2)) %>%
ungroup
# # A tibble: 6 x 5
# b1 b2 b3 CLASS OID
# <int> <int> <int> <chr> <int>
# 1 110 134 119 tree 1
# 2 112 133 118 tree 1
# 3 109 125 115 meadow 6
# 4 93 110 101 meadow 6
# 5 86 106 95 meadow 7
# 6 105 136 116 meadow 7
If your real data is arranged by OID like this example, you could also use top_frac
df %>%
group_by(CLASS) %>%
top_frac(.5, -OID)
# # A tibble: 6 x 5
# b1 b2 b3 CLASS OID
# <int> <int> <int> <chr> <int>
# 1 110 134 119 tree 1
# 2 112 133 118 tree 1
# 3 109 125 115 meadow 6
# 4 93 110 101 meadow 6
# 5 86 106 95 meadow 7
# 6 105 136 116 meadow 7

Answer 2

Your data:
df = structure(list(b1 = c(110L, 112L, 105L, 112L, 109L, 93L, 86L,
105L, 102L, 108L, 113L), b2 = c(134L, 133L, 125L, 132L, 125L,
110L, 106L, 136L, 128L, 129L, 134L), b3 = c(119L, 118L, 110L,
117L, 115L, 101L, 95L, 116L, 111L, 115L, 119L), CLASS = structure(c(2L,
2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = c("meadow",
"tree"), class = "factor"), OID = c(1L, 1L, 2L, 2L, 6L, 6L, 7L,
7L, 8L, 8L, 8L)), class = "data.frame", row.names = c(NA, -11L
))
First create a function to take 1/2 according to OID
func = function(x){
x[x$OID %in% x$OID[1:round(nrow(x)/2)],]
}
We randomize the way the OID are sorted
df$OID = factor(df$OID,levels=sample(unique(df$OID)))
df = df[order(df$OID),]
do.call(rbind,by(df,df$CLASS,func))
This will ensure you get random ~ 50% everytime, with complete OID