How to add multiple columns to a dataframe from a custom function in R

Question

I've created code that will take an input vector, create a dataframe based on the input, optimise some values and return some of these values. I'm now turning this into a function that will apply the calculations rowwise on an input dataframe. Below is a minimum working example of what I would like to achieve (my actual function would be too long to share here!):

# Randomly generated dataframe
df <-  data.frame(a = rnorm(10, 0, 1), x = rnorm(10, 1, 3), y = rnorm(10, 2, 3))

# Function that takes multiple arguments and returns multiple values in a list
zsummary <- function(x, y) { 
  if (y < 0) return(list(NA, NA))
  z = rnorm(10, x, abs(y))
  return(list(mean(z), sd(z)))
}

# Example of something that works using dplyr
#    However, this results in a lot of function calls...
#    especially if there were a lot of columns in the list...
library(dplyr)
df %>% rowwise() %>%
  mutate(mean = zsummary(x,y)[[1]], sd = zsummary(x,y)[[1]])

As you can see, I can't apply individual functions to each new df$mean and dfsd columns as they depend on a z vector that can only be generated once. I've looked around on SO already, but I haven't been able to find an answer yet. I think a solution would be using one of the apply functions and not something from dplyr, but I've honestly never fully understood apply functions. I would also not like solutions that use for loops with rbind as I've tried this in previous projects and for large dataframes it becomes very slow!

akrun · Accepted Answer · 2016-08-23 06:48:41Z

3

We can use mapply for this. As the zsummary takes two arguments, the mapply would be one option as it take corresponding element of 'x' and 'y' to apply the zsummary.

t(mapply(zsummary, df$x, df$y))

We can also change the function slightly and get the output with dplyr

zsummary <- function(x, y) { 
   if (y < 0) return(data.frame(mean = NA, sd = NA))
   z = rnorm(10, x, abs(y))
   data.frame(mean = mean(z), sd = sd(z))
}

 df %>%
     rowwise() %>% 
     do(data.frame(., zsummary(.$x, .$y)))

Or as we discussed in the comments, instead of having the function taking multiple arguments, have a single argument and use apply with MARGIN=1 for applying it on each row.

zsummary2 <- function(v1){
      if(v1[2] < 0) return(c(mean = NA, sd = NA))
      z <- rnorm(10, v1[1], abs(v1[2]))
       c(mean = mean(v1), sd= sd(v1))
     }

t(apply(df[-1], 1, zsummary2))
#         mean        sd
# [1,]  1.403066 0.8757504
# [2,]  5.058188 5.1401507
# [3,]  4.288365 1.4194393
# [4,]  1.932829 6.7587054
# [5,] -1.864236 3.7587462
# [6,]        NA        NA
# [7,]  3.328629 1.3711950
# [8,] -2.347699 5.0449958
# [9,]  2.936615 1.7332283
#[10,]        NA        NA

NOTE: The values will be different in each run as we didn't set any seed for the rnorm.

edited Aug 23, 2016 at 6:48

answered Aug 23, 2016 at 6:24

akrun

891k38 gold badges590 silver badges700 bronze badges

Sign up to request clarification or add additional context in comments.

6 Comments

Alwin Over a year ago

Thankyou for your quick reply! Would there be other ways to address x and y? My actual input has a dozen or so columns, so using something like colnames would be awesome to save me typing out input$a, input$b, input$c,... every time I want to use my function in mapply

akrun Over a year ago

@Alwin You can create the function in such a way that it takes a single argument like columns of dataset and then do the processing inside. It might be better.

Alwin Over a year ago

+1 also, since I don't have enough rep to upvote your answer. If there's nothing better I will accept it as the correct answer

akrun Over a year ago

@Alwin Thank you for the comments.

Alwin Over a year ago

+10 if I could, excellent answer! You would have saved me at least a couple hours of googling and trial and error. I've marked your answer as the accepted correct one

|

Collectives™ on Stack Overflow

How to add multiple columns to a dataframe from a custom function in R

1 Answer 1

6 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

6 Comments

Your Answer

Sign up or log in

Post as a guest

Related