tidyselect-friendly row-wise aggregation

Wrappers around base::pmin(), base::pmax(), lighthouse::psum(), and lighthouse::pmean() that accept tidyselect expressions.

Usage

psum_across(..., na.rm = FALSE)

pmean_across(..., na.rm = FALSE)

pmin_across(..., na.rm = FALSE)

pmax_across(..., na.rm = FALSE)

Arguments

...

<tidy-select> one or more tidyselect expressions that capture numeric and/or logical columns.

na.rm

Should missing values (including NaN) be removed?

Details

Lighthouse includes two sets of functions for computing "parallel" or row-wise aggregates:

• psum() and pmean() (which complement base::pmin() and pmax())

• pmin_across(), pmax_across(), psum_across(), and pmean_across()

Both sets of functions differ from base::rowSums() and rowMeans() in that they:

• work in data-masking contexts (e.g., inside dplyr::mutate()) without needing helpers like dplyr::pick() or dplyr::across().

• accept multiple inputs via ....

• return NA when na.rm = TRUE and all values in a row are NA. This mirrors behavior of base::pmin() and pmax(), but differs from rowSums(), which returns 0 in this situation.

psum_across() and friends support tidyselect expressions; e.g.,

dat 
  mutate(
    IDScrTotal = psum_across(IDScr1:IDScr6),
    SDScrTotal = psum_across(starts_with("SDScr"))
  )

...but must be used inside a data-masking verb like dplyr::mutate(), group_by(), or filter(), and do not support implicit computations.

Conversely, psum() and friends do not support tidyselect expressions, but can be used both inside or outside a data-masking context:

# data-masking
dat 
  mutate(
    NumColors = psum(Red, Blue, Green),
  )

#non-data masking
psum(1:10, 6:15, 11:20)

and support "on the fly" or "implicit" computations:

dat 
  mutate(
    CriteriaMet = psum(a1 > a2, pmax(b1, b2, b3) > 5, c1 < 10),
    TotalProp = psum(num1 / denom1, num2 / denom2, num3 / denom3)
  )

See also

psum() and pmax() for non-tidyselect parallel aggregates.

Examples

dat <- tibble::tribble(
  ~product,    ~price1, ~price2, ~price3,
  "Product 1", 20,      25,      22,
  "Product 2", NA,      30,      29,
  "Product 3", 15,      NA,      NA,
  "Product 4", NA,      NA,      NA
)

price_cols <- c("price1", "price2", "price3")

dat %>%
  dplyr::mutate(
    min = pmin_across(price1, price2, price3, na.rm = TRUE),
    max = pmax_across(price1:price3, na.rm = TRUE),
    sum = psum_across(starts_with("price"), na.rm = TRUE),
    mean = pmean_across(all_of(price_cols), na.rm = TRUE)
  )
#> # A tibble: 4 × 8
#>   product   price1 price2 price3   min   max   sum  mean
#>   <chr>      <dbl>  <dbl>  <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 Product 1     20     25     22    20    25    67  22.3
#> 2 Product 2     NA     30     29    29    30    59  29.5
#> 3 Product 3     15     NA     NA    15    15    15  15  
#> 4 Product 4     NA     NA     NA    NA    NA    NA  NA