Search code examples
rrandomsampling

R row-wise sampling with upside down normal distribution

I have a data frame DF that looks like this:

Min Max
10  80
20  90
40  120

I want to append a new Random column with a randomly generated number between the min and max values per row. The sampling of the number shall be derived from an upside down normal distribution and exclude the middle values like illustrated below

enter image description here

The below prototype code with single variables seem to work but I'm stuck with how to apply this row-wise.

min <- 1
max <- 20
q <- min + (max-min)*rbeta(10000, 0.5, 0.5)
q <- q[!(q > 5 & q < 15)][1:10000]
hist(q)

enter image description here

Solution

  • You could try this iterative approach. Use the variables lower and upper to define the excluded middle range.

    Start by creating a column of NA values. For each iteration of the loop, all NA values in the column are overwritten with samples from your distribution. The samples that are within the excluded zone are then overwritten with NA and the loop repeats until no NA values are left in the column.

    DF    <- data.frame(Min = c(10, 20, 40), Max = c(80, 90, 120))
lower <- 5
upper <- 15

DF$sample <- rep(NA, nrow(DF)); 
while(any(is.na(DF$sample))) 
{ 
  i <- which(is.na(DF$sample)); 
  DF$sample[i] <- DF$Min[i] + (DF$Max[i] - DF$Min[i]) * rbeta(length(i), 0.5, 0.5); 
  DF$sample[DF$sample > lower & DF$sample < upper] <- NA;
}

DF
#>   Min Max   sample
#> 1  10  80 31.88867
#> 2  20  90 33.26248
#> 3  40 120 80.08321

    Created on 2020-02-18 by the reprex package (v0.3.0)