I have a dataset like this
alpha number fr color
1 a 20 0.8 rot
2 a 21 2.0 rot
3 a 2 0.8 rot
4 a 34 0.8 rot
5 f 42 0.5 grün .......
......................
now i want to split this dataset in more observations which depends on a condition like number<20 so that the new dataset looks like
alpha number fr color
1 a 19 0.8 rot
2 a 1 0.8 rot
3 a 10 2.0 rot
4 a 11 2.0 rot
5 a 2 0.8 rot
6 a 19 0.8 rot
7 a 15 0.8 rot
8 f 7 0.5 grün
9 f 7 0.5 grün
10 f 7 0.5 grün
11 f 7 0.5 grün
12 f 7 0.5 grün
13 f 7 0.5 grün
.......
or similar. repeat the observations as long as the condition is not true. how to split doesnt matter but the observations have to be the same for the data you split for the other variables.
thank you
df1 <- structure(list(alpha = c("a", "a", "a", "a", "f"),
number = c(20L, 21L, 2L, 34L, 42L),
fr = c(0.8, 2, 0.8, 0.8, 0.5),
color = c("rot", "rot", "rot", "rot", "grun")),
row.names = c(NA, -5L), class = "data.frame")
rep.rev <- function(x,t){
if(t != 0){
rep(x,t)
} else {
NA_integer_
}
}
library(dplyr)
library(tidyr)
set.seed(22)
df1 %>%
mutate(divisor = floor(runif(n(), min = 2, max = 19)),
quotient = number%/%divisor,
remainder = ifelse(number%%divisor==0, NA, number%%divisor)) %>%
rowwise %>%
mutate(number = list(c(rep.rev(divisor, quotient),remainder))) %>%
unnest %>%
select(alpha, number, fr, color) %>%
filter(!is.na(number))
#> # A tibble: 14 x 4
#> alpha number fr color
#> <chr> <dbl> <dbl> <chr>
#> 1 a 7 0.8 rot
#> 2 a 7 0.8 rot
#> 3 a 6 0.8 rot
#> 4 a 10 2 rot
#> 5 a 10 2 rot
#> 6 a 1 2 rot
#> 7 a 2 0.8 rot
#> 8 a 10 0.8 rot
#> 9 a 10 0.8 rot
#> 10 a 10 0.8 rot
#> 11 a 4 0.8 rot
#> 12 f 16 0.5 grun
#> 13 f 16 0.5 grun
#> 14 f 10 0.5 grun
Many ways we could split a number >=20, but the method below divides the each number in half, i.e. even numbers (m=2n) split into n and n, odd numbers (m=2n+1) split into (n+1) and n.
> library(dplyr)
> df <- data.frame(alpha=c("a","a","a","a","f"),
+ number=c(20,21,2,34,42),
+ fr=c(0.8,2.0,0.8,0.8,0.5),
+ color=c("rot","rot","rot","rot","grün"))
The function doSplit() takes a dataframe df and an integer threshold as arguments.
> doSplit <- function(df, threshold){
+ # splits rows where number >= threshold until all rows have number < threshold
+
+ colNames <- colnames(df)
+ df <- df %>% mutate(orig_id=rownames(df))
+ dfBelow <- df %>% filter(number<threshold)
+ dfAbove1 <- df %>% filter(number>=threshold) %>% mutate(number=(number%/%2)+(number%%2))
+ dfAbove2 <- df %>% filter(number>=threshold) %>% mutate(number=number%/%2)
+ combData <- rbind(dfBelow, dfAbove1, dfAbove2)
+ combData <- combData %>% arrange(orig_id) %>% select(colNames)
+ return(combData)
+ }
Here, we define a threshold of 20. The while loop repeatedly calls the doSplit() function so long as a row exists with number >=20.
> myThreshold <- 20
> splitDf <- df
> while(splitDf %>% pull(number) %>% max() >= myThreshold){
+ splitDf <- doSplit(splitDf, myThreshold)
+ }
Here is the split dataframe:
> splitDf
alpha number fr color
1 a 10 0.8 rot
2 a 10 0.8 rot
3 a 11 2.0 rot
4 a 10 2.0 rot
5 a 2 0.8 rot
6 a 17 0.8 rot
7 a 17 0.8 rot
8 f 11 0.5 grün
9 f 10 0.5 grün
10 f 11 0.5 grün
11 f 10 0.5 grün
