Access column name inside function used inside summarize_all dplyr structure - r

Question

I'm building a dplyr structure to run some custom functions over the columns of a dataframe in 1 block of code
currently my function looks this
funx <- function(x) {
logchoice <- if(max(x) < 400) {'T' } else { 'F' }
logtest <- suppressWarnings(log10(x))
remaining <- length(logtest[which(!is.na(logtest) & is.finite(logtest))])
x <- if(remaining > 0.75*length(x)) {suppressWarnings(log10(x)) } else { x }
x <- x[which(!is.na(x) & is.finite(x))]
y <- diptest::dip.test(x)
z <- tibble(pvalue = y$p.value, Transform = logchoice)
return(z)
}
and the dplyr structure looks like this:
mtcars %>%
sample_n(30) %>%
select(colnames(mtcars)[2:5]) %>%
summarise_all(list(~ list(funx(.)))) %>%
gather %>%
unnest %>%
arrange(pvalue) %>%
rename(Parameter = key)
which gives me:
Parameter pvalue Transform
1 cyl 0.00000000 T
2 drat 0.03026093 T
3 hp 0.04252001 T
4 disp 0.06050505 F
I would like to know how I can access the column name inside my function, mainly because I would like to change the name in the result table to look like the output of this: paste(original_column_name, 'log10', sep = '') if the function applies the log transformation, but leave the original name as is when it decides not to.
so the expected output is:
Parameter pvalue Transform
1 log10_cyl 0.00000000 T
2 log10_drat 0.03026093 T
3 log10_hp 0.04252001 T
4 disp 0.06050505 F

Answer 1

You were quite close. You can just add a mutate() to the end
mtcars %>%
sample_n(30) %>%
select(colnames(mtcars)[2:5]) %>%
summarise_all(list(~ list(funx(.)))) %>%
gather() %>%
unnest() %>%
arrange(pvalue) %>%
rename(Parameter = key) %>%
mutate(Parameter = ifelse(Transform == "T", paste0("log10_", Parameter), Parameter)) %>%
select(Parameter, pvalue)
# Parameter pvalue
# log10_cyl 0.00000000
# log10_drat 0.01389723
# disp 0.02771770
# log10_hp 0.08493466

Answer 2

Answering in a separate post as the solution is a different. To get the column names in a print(), I would pass them in the function and use purrr::map_dfr to build a dataframe of the result. The small changes I made are to grab the column name, col_name, and specify the dataframe. I tried a few approaches to grab the column name using your original function but came out unsuccessful.
logtest_pval <- function(col, df) {
col_name <- col
x <- df %>% pull(!!col)
logchoice <- ifelse(max(x) < 400, TRUE, FALSE)
logtest <- log10(x)
remaining <- length(logtest[which(!is.na(logtest) & is.finite(logtest))])
x <- if(remaining > 0.75*length(x)) {suppressWarnings(log10(x)) } else { x }
x <- x[which(!is.na(x) & is.finite(x))]
y <- diptest::dip.test(x)
z <-
tibble(
transform = logchoice,
column = ifelse(logchoice, paste0("log10_", col_name), col_name),
pvalue = y$p.value
)
print(paste0(z, collapse = " | "))
return(z)
}
Then you can build your dataframe:
purrr::map_dfr(
.x = names(mtcars), # the columns to use
.f = logtest_pval, # the function to use
df = mtcars # additional arguments needed
)
Here's another example
df <-
mtcars %>%
select_if(is.numeric)
pvalues <-
map_dfr(names(df), logtest_pval, df)