Passing multiple arguments into map() with prop.test() with dplyr - r

Question

could anybody help my to change the rowwise() + prop.test() to map2?
The issue is, that map2 takes .x and .y arguments, but I need to pass 4 columns: 2xsucces and 2xtries into the prop.test() function. I want to extract the p-value into new column p.
This code runs OK'isch:
library(tidyverse)
cc <- tribble(
~n1, ~x1, ~n2, ~x2,
1000,100,900,85,
1000,100,100,10,
1000,100,10,10
)
cc %>%
rowwise()%>%
mutate(p = prop.test(x=c(x1, x2),
n=c(n1, n2),
conf.level=0.95)$p.value)%>%
mutate(p=round(p,5))
#> Warning in prop.test(x = c(x1, x2), n = c(n1, n2), conf.level = 0.95): Chi-
#> squared approximation may be incorrect
#> # A tibble: 3 × 5
#> # Rowwise:
#> n1 x1 n2 x2 p
#> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 1000 100 900 85 0.741
#> 2 1000 100 100 10 1
#> 3 1000 100 10 10 0
But I noticed that I can gain speed with a for loop:
cc <- tribble(
~n1, ~x1, ~n2, ~x2,
1000,100,900,85,
1000,100,100,10,
1000,100,10,10
)
cc %>%
rowwise()%>%
mutate(p = prop.test(x=c(x1, x2),
n=c(n1, n2),
conf.level=0.95)$p.value)%>%
mutate(p=round(p,5))
cc$p <-999
for (i in 1:nrow(cc)){
result <- prop.test(x=c(cc$x1[[i]], cc$x2[[i]]),
n = c(cc$n1[[i]], cc$n2[[i]]),
conf.level=0.95)
cc$p[[i]] <- result$p.value
}
cc %>%
mutate(p=round(p,5))
}
But I ask myself it there is more elegant way to use map function from dplyr in order to write less code and achieve the same results?
Thanks in advance

Answer 1

There is pmap:
library(tidyverse)
cc <- tribble(
~n1, ~x1, ~n2, ~x2,
1000, 100, 900, 85,
1000, 100, 100, 10,
1000, 100, 10, 10
)
cc %>%
mutate(
signif = list(x1, x2, n1, n2) %>% pmap_dbl(~ {
prop.test(
x = c(..1, ..2),
n = c(..3, ..4),
conf.level = 0.95
)$p.value
})
)
#> Warning in prop.test(x = c(..1, ..2), n = c(..3, ..4), conf.level = 0.95): Chi-
#> squared approximation may be incorrect
#> # A tibble: 3 x 5
#> n1 x1 n2 x2 signif
#> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 1000 100 900 85 7.41e- 1
#> 2 1000 100 100 10 1 e+ 0
#> 3 1000 100 10 10 9.46e-18
Created on 2022-02-22 by the reprex package (v2.0.0)
For loops are usually very slow, and the code is hard to read and often unnecessary when it comes to the scenario of just apply a function for each element.
The thing which slows down your dplyr code is rowwise, which is not needed anymore using the map function.