CODEWITHSUNDEEP
rcrosstab
wishihadabettername
wishihadabettername

Reputation: 14751

How to create contigency tables (crosstabs) in R for a subset of columns with categorical data?

I have a table whose header looks like this (I've simplified it):

id, a1, a2, a3, a4, a5, a6, a7, a8, a9, a10

where each row, except for id, is a categorical variable. Let's name the categories A, B, C, D, E.

I would like to create a contingency table for some of the columns, such as below (for brevity, I have not put sample numbers in the cells). Getting the total column/row would be great, but not mandatory, I can calculate it myself later.

      a1  a2  a3  a4 Total
    ----------------------
    A|
    B|
    C|
    D|
    E|
Total|

Thus, the question is how to create a crosstab based on multiple columns in R? The examples I've seen with table() and xtabs() use a column only. In my case, the columns are adjacent, so one crosstab would summarize columns a1..a4, another a5..a7 and so on. I hope there is an elegant way to do this.

I'm a programmer, but a newbie in R.

Thank you in advance.

Upvotes: 2

Views: 11806

Answers (2)

Joshua Ulrich
Joshua Ulrich

Reputation: 176718

Here's how to do it using base R commands. You don't need the for loop if every column has the same factor levels, but the loop would be a good fail-safe.

> set.seed(21)
> df <- data.frame(
+   id=1:20,
+   a1=sample(letters[1:4],20,TRUE),
+   a2=sample(letters[1:5],20,TRUE),
+   a3=sample(letters[2:5],20,TRUE),
+   a4=sample(letters[1:5],20,TRUE),
+   a5=sample(letters[1:5],20,TRUE),
+   a6=sample(letters[1:5],20,TRUE) )
> 
> for(i in 2:NCOL(df)) {
+   levels(df[,i]) <- list(a="a",b="b",c="c",d="d",e="e")
+ }
> 
> addmargins(mapply(table,df[,-1]))
    a1 a2 a3 a4 a5 a6 Sum
a    6  2  0  2  5  3  18
b    3  3  7  2  1  3  19
c    5  3  1  6  5  3  23
d    6  8  6  1  5  3  29
e    0  4  6  9  4  8  31
Sum 20 20 20 20 20 20 120

Upvotes: 3

JoFrhwld
JoFrhwld

Reputation: 9047

Your data is poorly formatted for this purpose. Here's one approach to appropriately reshaping the data with the reshape package.

library(reshape)
data.m <- melt(data, id = "id")

To compute a table for all levels, with margins, you could use

cast(data.m, value ~ variable, margins = T)

For a subset, take the relevant subset of data.m.

Upvotes: 7

Related Questions