Complex data reshaping in R

Question

I have a data frame with 3 columns (extract below):

df <- data.frame(
  id = c(1,1,1,2,2,2),
  Year = c(2007, 2008, 2009, 2007, 2008, 2009),
  A = c(5, 2, 3, 7, 5, 6),
  B = c(10, 0, 50, 13, 17, 17) 
)
df  

I'd like to have this:

df_needed <- data.frame(
  id= c(1, 2),
  A_2007 = c(5, 7),
  B_2007 = c(10, 13),
  A_2008 = c(2, 5),
  B_2008 = c(0, 17),
  A_2009 = c(3, 6),
  B_2009 = c(50, 17)
)
df_needed

I'm familiar with reshape and tidyR but I don't think they can manage this transformation.

Is there a proper way to do that or I need to do it with a custom function ?

Edit : This example has been edited in order to improve the example with more than 1 record in the final data set.

Answer 1

Try

library(dplyr)
library(tidyr) 
gather(df, Var, Val, -Year) %>% 
        unite(YearVar, Var, Year)  %>% 
        mutate(indx=1) %>% 
        spread(YearVar, Val)%>%
        select(-indx)
 #  A_2007 A_2008 A_2009 B_2007 B_2008 B_2009
 #1      5      2      3     10      0     50

Update

For the edit, you can change the variables in the gather

gather(df, Var, Val, A:B) %>% 
         unite(YearVar, Var, Year) %>% 
         spread(YearVar, Val)
#   id A_2007 A_2008 A_2009 B_2007 B_2008 B_2009
#1  1      5      2      3     10      0     50
#2  2      7      5      6     13     17     17

Answer 2

Here's a possible solution using data.table v >= 1.9.5

library(data.table)
dcast(setDT(df), . ~ Year, value.var = c("A", "B"))
#    . 2007_A 2008_A 2009_A 2007_B 2008_B 2009_B
# 1: .      5      2      3     10      0     50

---

Edit : per your new data set, simply add id to the formula

dcast(setDT(df), id ~ Year, value.var = c("A", "B"))
#    id 2007_A 2008_A 2009_A 2007_B 2008_B 2009_B
# 1:  1      5      2      3     10      0     50
# 2:  2      7      5      6     13     17     17

Answer 3

Good ol' base::reshape works fine here. Just create a dummy id variable first.

df$id <- 1
reshape(df, v.names = c("A", "B"), timevar = "Year", idvar = "id", direction = "wide")
#   id A.2007 B.2007 A.2008 B.2008 A.2009 B.2009
# 1  1      5     10      2      0      3     50

To save some typing, given that you specify timevar and idvar , you don't have to provide v.names :

reshape(df, timevar = "Year", idvar = "id", direction = "wide")

This works for the edited data as well (which happened to have the "id" variable already).

#    id A_2007 B_2007 A_2008 B_2008 A_2009 B_2009
#  1  1      5     10      2      0      3     50
#  2  2      7     13      5     17      6     17

You may also use reshape2::recast :

recast(df, id ~ variable + Year, id.var = 1:2)

Answer 4

Another simple option in base R :

df_needed <- matrix(as.vector(t(df[, -1])), ncol=nrow(df)*(ncol(df)-1))
colnames(df_needed) <- paste(rep(colnames(df)[-1], nrow(df)), rep(df[, 1], e=ncol(df)-1), sep="_")

df_needed
#      A_2007 B_2007 A_2008 B_2008 A_2009 B_2009
#[1,]      5     10      2      0      3     50

with edited data

df_split <- split(df, df$Year)
df_split <- lapply(df_split, function(df){colnames(df)[-1] <- paste(colnames(df)[-1], unique(df$Year), sep="_"); df <- df[, -1]; return(df)})
df_needed <- do.call("cbind", df_split)
colnames(df_needed) <- sub("^\\d{4}\\.","",colnames(df_needed))
df_needed
#  A_2007 B_2007 A_2008 B_2008 A_2009 B_2009
#1      5     10      2      0      3     50
#4      7     13      5     17      6     17