Vectorized conditional outputs in R (maybe using purrr)?

user_8301930 0
Viewed 83

I am trying to create vectorized conditional outputs of a data frame.

Suppose I have the dataframe:

data <- data.frame(a = c(5, 3, 9, 5),
                   b = c(1, 2, 3, 4),
                   c = c(5, 3, 9, 5),
                   d = c(5, 3, 9, 5))

And the threshold:

threshold <- c(a1 = 4, b1 = 2, c1 = 8, d1 = 2)

What I want is a new dataset that indicates whether each value of a is greater than or equal to the corresponding value in the threshold vector (a1), each value of b is greater than or equal to the corresponding value in the threshold vector (b1), etc.

So the desired output would be:

desired_data <- data.frame(a = c(1, 0, 1, 1),
                           b = c(0, 1, 1, 1),
                           c = c(0, 0, 1, 0),
                           d = c(1, 1, 1, 1))

I want to do this as simply as possible, ideally using a purrr function.

Here is a wrong attempt:

desired_data <- map(data >= threshold)

I feel like map2 might be promising, and have checked the documentation (e.g., here and here) but I can't seem to get the syntax for conditional outputs based on mapping.

Thank you!

purrr conditional-statements r
5 Answers

You have the idea with map2(): since a data frame is a list of column, you can loop on the columns. The small difficulty is to put everything in a data frame at the end, it's done automatically if you use map2_df().

map2_df(threshold, data, ~ .y >= .x)

And if you want these logical values to be converted to integers:

1L * map2_df(threshold, data, ~ .y >= .x)
#   a1 b1 c1 d1
# 1  1  0  0  1
# 2  0  0  0  1
# 3  1  1  1  1
# 4  1  1  0  1
user_39383600

Another option is sweep in base R. However, the output is a matrix and has to be converted to data.frame:

as.data.frame(1L * sweep(data, 2, threshold, FUN=">="))
#   a b c d
# 1 1 0 0 1
# 2 0 1 0 1
# 3 1 1 1 1
# 4 1 1 0 1
user_39265430

Using pmap from the purrr package.

library(purrr)

data <- data.frame(a = c(5, 3, 9, 5),
                   b = c(1, 2, 3, 4),
                   c = c(5, 3, 9, 5),
                   d = c(5, 3, 9, 5))

desired_data <- data.frame(a = c(1, 0, 1, 1),
                           b = c(0, 1, 1, 1),
                           c = c(0, 0, 1, 0),
                           d = c(1, 1, 1, 1))

threshold <- c(a1 = 4, b1 = 2, c1 = 8, d1 = 2)

output <- pmap(list(data, threshold), ~ .x >= .y)

as.data.frame(1L * do.call(cbind, output))

#>   a b c d
#> 1 1 0 0 1
#> 2 0 1 0 1
#> 3 1 1 1 1
#> 4 1 1 0 1

Created on 2020-11-17 by the reprex package (v0.3.0)

user_133567300

you can use replicate to create a matrix with each row containing the the thresholds: 1st row has as .... then transpose and compare it with data, make it a data.frame then add 0 to convert the values to numeric :

data.frame( data >= t(replicate(nrow(data), threshold)) ) + 0

  a b c d
1 1 0 0 1
2 0 1 0 1
3 1 1 1 1
4 1 1 0 1
user_39808200

Another base R option :

+t(t(data) >= threshold)

#     a b c d
#[1,] 1 0 0 1
#[2,] 0 1 0 1
#[3,] 1 1 1 1
#[4,] 1 1 0 1
user_39629140
Related