Question

I am trying to reshape my data from this form in exhibit A to the form in exhibit B. I’ve tried reshape and looping over the data by each three columns and appending the datasets, but can’t quite get there. How can I reshape this?

A   AFG1    AFG2    AFG3    US1 US2 US3 t
    5   7   9   3   4   5   1980
    6   8   10  4   6   5   2000


B   1   2   3   t   xtry        
    5   7   9   1980    AFG     
    6   8   10  2000    AFG     
    3   4   5   1980    US      
    4   6   5   2000    US      

Answer 1

Assuming your data.frame is called “mydf”, try this:

reshape(mydf, direction = "long",
        idvar = "t",
        varying = 1:6,
        v.names = c("1", "2", "3"), 
        times = c("AFG", "US"),
        timevar = "xtry")
#             t xtry 1 2  3
# 1980.AFG 1980  AFG 5 7  9
# 2000.AFG 2000  AFG 6 8 10
# 1980.US  1980   US 3 4  5
# 2000.US  2000   US 4 6  5

---

For your problem, you have to be a little more verbose in your arguments than a straightforward reshape problem because your variables are named differently than what R expects (which is in the form of 1.AFG, 2.AFG, and so on).

For example:

mydf2 <- mydf
names(mydf2) <- gsub("([A-Z]+)([0-9]+)", "\\2\\.\\1", names(mydf2))

If your names looked like the following:

names(mydf2)
# [1] "1.AFG" "2.AFG" "3.AFG" "1.US"  "2.US"  "3.US"  "t" 

The reshape command is a bit more direct.

reshape(mydf2, direction = "long", idvar = "t",
        timevar = "xtry", varying = 1:6)

---

reshape2

If you are looking for a “reshape2” solution, I actually find it a little more work than using base R’s reshape function (it’s usually the other way around). Here’s what I came up with:

First, melt the dataset.

library(reshape2)
mydf_m <- melt(mydf, id.vars="t")
head(mydf_m)
#      t variable value
# 1 1980     AFG1     5
# 2 2000     AFG1     6
# 3 1980     AFG2     7
# 4 2000     AFG2     8
# 5 1980     AFG3     9
# 6 2000     AFG3    10

That “variable” column is pretty useless to us in its current form, so let’s fix it.

mydf_m <- cbind(mydf_m, 
                colsplit(gsub("([A-Z]+)([0-9]+)", 
                              "\\1_\\2", mydf_m$variable), 
                         "_", c("xtry", "var")))

Here’s what the data look like now.

head(mydf_m)
#      t variable value xtry var
# 1 1980     AFG1     5  AFG   1
# 2 2000     AFG1     6  AFG   1
# 3 1980     AFG2     7  AFG   2
# 4 2000     AFG2     8  AFG   2
# 5 1980     AFG3     9  AFG   3
# 6 2000     AFG3    10  AFG   3

Now comes the easy part.

dcast(mydf_m, t + xtry ~ var)
#      t xtry 1 2  3
# 1 1980  AFG 5 7  9
# 2 1980   US 3 4  5
# 3 2000  AFG 6 8 10
# 4 2000   US 4 6  5