How can you reference non-selected columns within across()?

Question

I have a dataframe with integer columns x1, x2, x3, ... and bool columns x1_status, x2_status, x3_status, ...

df = tibble(
    x1 = 1:5,
    x1_status = c(T, T, T, F, F),
    x2 = 6:10,
    x2_status = c(F, T, T, T, F),
    x3 = 11:15,
    x3_status = c(F, F, T, T, T)
)

How do I calculate the mean of x1, x2, x3, ... at the values where x1_status is True, x2_status is True, x3_status is True, and so on, respectively? I'd like to use across() within summarize() so I don't have to list each x column individually.

That is, I'm hoping to get a dataframe with columns mean_x1 = 2, mean_x2 = 8, and mean_x3 = 14.

I was thinking of something like

df |>
    summarize(
        across(
            !matches("_"),
            \(value_col) {
                status_col = str_c(cur_column(), "_status")
                mean(value_col[.data[[status_col]]])
            },
            .names = "mean_{col}"
        )
    )

but that doesn't work. I'd also like to avoid using df[[status_col]], so I can include this in a pipeline that might create the x1, x2, x3, ... columns without having to assign df in the interim.

Answer 1

Using pivot_longer() might be more intuitive:

library(tidyverse)

df = tibble(
  x1 = 1:5,
  x1_status = c(T, T, T, F, F),
  x2 = 6:10,
  x2_status = c(F, T, T, T, F),
  x3 = 11:15,
  x3_status = c(F, F, T, T, T)
)

# Pivot to longer dataset
df_long <- df %>%
  rename_with(
    ~ paste0(., "_value"),
    c(x1, x2, x3)
  ) %>%
  pivot_longer(
    cols = everything(),
    names_sep = "_",
    names_to = c("name", ".value")
  )

# Summarize
long_summary <- df_long %>%
  filter(status) %>%
  group_by(name) %>%
  summarize(value = mean(value))

# Pivot to wide summary
wide_summary <- long_summary %>%
  pivot_wider(
    names_from = "name",
    values_from = "value",
    names_prefix = "mean_"
  )

print(wide_summary)