在按计数汇总之前折叠列

Question

I have what I imagine is a very simple question, but I simply cannot figure it out after a good deal of board searching/reading tutorials.

I have a df with name entries in columns 5 through 12 that are all of one type. They are name strings (see below for example). All I would like to do is use the aggregate or ddply (or another, if easier...) function to collapse those columns and then return the count of each unique entry.

ID | Name 1 | Name 2 | Name 3 
Row 1: 278 | John | Tim | Mike
Row 2: 279 | Tim | Steve | John
Row 3: 280 | Tim | Doug | Dave 

So ideally I'd get:

 Tim | 3 
 John | 2
 Mike | 1 
 etc. | 1 

I know how this works for one column:

counts=aggregate(numeric(nrow(df)), df[c(4)], length)

But when I use a similar line for multiple columns, it returns the unique combinations of the seven columns, instead of an nx2 vector with the aggregated unique entries and total sums.

counts2=aggregate(numeric(nrow(df)),df[c(5:12)],FUN = function(x) length(unique(x)))

Thank you very much for your help.

Answer 1

Here is one way using dplyr and tidyr

foo <- data.frame(id = 278:280,
                  Name1 = c("John", "Tim", "Mike"),
                  Name2 = c("Tim", "Steve", "John"),
                  Name3 = c("Tim", "Doug", "Dave"),
                  stringsAsFactors = FALSE)
library(dplyr)
library(tidyr)

foo %>%
    gather(var, names, -id) %>%
    count(names)

#  names n
#1  Dave 1
#2  Doug 1
#3  John 2
#4  Mike 1
#5 Steve 1
#6   Tim 3

Answer 2

I'm not as up to speed on the new packages that Hadley has come up with, but here's how I'd solve the problem using reshape2 package. The idea (same as above) is to collapse the columns into one column and then summarize that data:

library(reshape2)

dcast(data = melt(foo, id.vars = "id"), value ~ .)
#---
  value .
1  Dave 1
2  Doug 1
3  John 2
4  Mike 1
5 Steve 1
6   Tim 3

Answer 3

Reading your data:

txt <- "ID | Name 1 | Name 2 | Name 3 
Row 1: 278 | John | Tim | Mike
Row 2: 279 | Tim | Steve | John
Row 3: 280 | Tim | Doug | Dave "
dat <- read.csv(text = txt, sep = "|", strip.white = TRUE)

You can use the as.data.frame table method on the unlisted columns.

u <- unlist(dat[-1])
as.data.frame(table(levels(u)[u]))
#    Var1 Freq
# 1  Dave    1
# 2  Doug    1
# 3  John    2
# 4  Mike    1
# 5 Steve    1
# 6   Tim    3