有没有一种方法可以用运行速度更快的方法替换R中的此循环？

Question

To make it short, I am trying to speed things up. This is my slow code:

library(dplyr)
tmp <- unique(kat$pnr) # Sort out the unique entries (ends up to about 572000)
sex = c()
for(i in tmp){         # For each unique pnr, look up the sex and append it to the new dataset
  temptable <- filter(kat, pnr == i)
  sex[i] <- temptable$sex
}

Currently the loop will take me hours as I have 572000 rows to loop through in the tmp-dataset and it seems like the system processes about 50 rows a second when I do some shorter test runs. So is there a way to replace this loop with something that runs faster?

In the kat-dataset I have about 40 columns and 905000 rows of which the pnr is the unique identifier, however, one pnr can occur one or two times in kat. I want to do gender statistics so I basically want to sort out the unique pnrs and the sex of each pnr.

Answer 1

The == from unique elements with filter would be slower and that too on a loop. Instead, for this case, a groupby operation may be more appropriate if we want to find some descriptive statistics on the 'sex' column for each unique element of 'pnr'

library(dplyr)
kat %>%
    group_by(pnr) %>%
    summarise(val = fn(sex))

---

It can be made further faster with data.table

library(data.table)
setDT(kat)[, .(val = fn(sex)), by = .(pnr)]

NOTE: not clear about the function to apply on the 'sex' column

---

If the intention is to create a list of sex , then

lst1 <- split(kat$sex, kat$pnr)