How to Find Maximum Absolute Values by Row in Data Frame in R

This article will explain how to find maximum absolute values by row in a data frame in R.

Use Custom Function to Find Maximum Absolute Values by Row in R Data Frame

Finding maximum absolute values in each row of the data frame and constructing them as a vector can be done using the abs_max function that we implemented in the following code snippet. We assume the data frame contains numeric and non-numeric values and initialize an object called df1 for verification purposes. The first function called in the abs_max block is Filter. It takes two arguments. The first represents the unary function applied to each element of the vector passed as the second argument, and Filter extracts the elements that evaluate true. In this case, Filter takes the is.numeric function to extract only numeric values from the data frame.

Then, the inherits function is utilized to check if the argument of the abs_max is of type tbl_df, and if so, it is converted to a matrix. Note that, tbl_df class is commonly known as tibble and is provided with the same name by a separate package, which we will utilize in the next example. The replace function is called to replace negative values with their absolute value and the NA types with -Inf. The next, max.col is invoked and extracts the vector of positions for the maximum values in each row. Finally, the result values are taken using the cbind function. The latter object we can append to the existing data frame as a separate column named abs_max as demonstrated in the following code sample.

abs_max <- function(data) {
  tmp <- Filter(is.numeric, data)
  if(inherits(data, "tbl_df")) {
    tmp <- as.matrix(tmp)
  }
  tmp[cbind(1:nrow(tmp), max.col(replace(x <- abs(tmp), is.na(x), -Inf)))]
}

df1 <- data.frame(
  id = c("foo", "bar", "goo"),
  val_a =  c(-51, 15, 19),
  val_b = c(NA, 122, 35),
  val_c = c(10, -23, 4)
)

df1$abs_max = abs_max(df1)
df1

Output:

id val_a val_b val_c abs_max
1 foo   -51    NA    10     -51
2 bar    15   122   -23     122
3 goo    19    35     4      35

Alternatively, the same function can be used to pass the tibble object type and process its rows. Note that tibbles will be converted to matrices in the abs_max function, but the rest of the function code works the same way it did on the data frame.

library(tibble)

abs_max <- function(data) {
  tmp <- Filter(is.numeric, data)
  if(inherits(data, "tbl_df")) {
    tmp <- as.matrix(tmp)
  }
  tmp[cbind(1:nrow(tmp), max.col(replace(x <- abs(tmp), is.na(x), -Inf)))]
}

tb1 <- tibble(
  id = c("foo", "bar", "goo"),
  val_a =  c(-51, 15, 19),
  val_b = c(NA, 122, 35),
  val_c = c(10, -23, 4)
)

tb1$abs_max = abs_max(tb1)
tb1

Output:

# A tibble: 3 x 5
  id    val_a val_b val_c abs_max
  <chr> <dbl> <dbl> <dbl>   <dbl>
1 foo     -51    NA    10     -51
2 bar      15   122   -23     122
3 goo      19    35     4      35