Why is my subsampling extremely biased for high quality cells?

[markphillippebworth]

Hello! I don't know if this is better submitted as a bug or discussion question.

I've been running some reproducibility analysis by subsampling one particular cell type from one sample. I wrote my code so that the sampling should be random from the population, with ever increasing number of cells selected. However, when I look at the number of nFrags per cell, it looks the fewer cells I choose, the more biased they were towards have more nFrags per cells. My overall dataset has an average of about 4000 fragments per cell.

Here's what it looks like across iterations.

Below is my code for subsetting.

cellRange <- c(seq(10,110, by = 20),seq(150,500, by = 50),seq(600,2400, by = 100))
#[1]   10   30   50   70   90  110  150  200  250  300  350  400  450  500  600
#[16]  700  800  900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100
#[31] 2200 2300 2400

for(i in 1:length(cellRange)){
CD4sets <- subsetArchRProject(CD4Test, getCellNames(CD4Test)[sample(cellRange[i]*2+22)], outputDirectory = "CD4sets/", force = TRUE)
saveArchRProject(CD4sets, "CD4sets/")
}

[rcorces]

Maybe its your (mis)usage of sample?
I believe you are supposed to provide a vector of values to sample from. so it should be
sample(x = c(1:length(cellNames(CD4Test)), size = cellRange[i]*2+22)
Otherwise, with the way you are using it, you randomly sample (without replacement) the numbers from 1 to cellRange[i]*2+22

for ex.

> sample(5)
[1] 4 2 1 5 3

[markphillippebworth]

yep, that would do it! Thank you.