0

I was wondering how I might simply split a numerical column by a second grouping variable in a dataset, then cbind the numerical column. This would most likely be a simple extension of the separate function for dplyr. For example, changing X below: 

Y <- rbind(2,5,3,6,3,2)      
Z <- rbind("A", "A", "A", "B", "B", "B")   
X <- data.frame(Y,Z)

Into 

  A B                   
  2 6      
  5 3    
  3 2

Then ideally extract the rowMeans into a new vector. (Issue also arises here when there is only one character in Z, given rowmeans requires 2).

This would need to be infinitely expandable based on the number of unique variables in Z. e.g., if Z had A, B, and C, then the final data.frame would require 3 columns. This would allow me to capture the row means from infinite number of groups in Z.

Thanks in advance, Conal

www
  • 38,575
  • 12
  • 48
  • 84
Conal Monaghan
  • 87
  • 7

1 Answers1

1

Looks like a job for tidyr::spread. 

library(dplyr)
library(tidyr)

X2 <- X %>%
  group_by(Z) %>%
  mutate(ID = 1:n()) %>%
  spread(Z, Y) %>%
  select(-ID)
X2
# A tibble: 3 x 2
      A     B
* <dbl> <dbl>
1     2     6
2     5     3
3     3     2
www
  • 38,575
  • 12
  • 48
  • 84