CODEWITHSUNDEEP
rcontingency
iNyar
iNyar

Reputation: 2326

Creating a contingency table from multiple logical variables in R

I currently have a data frame with one factor and multiple logical variables (that correspond to some extent to various conditions of a single variable, but are NOT excluding one another).

Taking a simplistic comparison, I want to count cars from different types of people, but someone can have various cars. And let's say I also want to count their phone type (also not excluding one another).

Dummy data:

data = data.frame(Profession = c("Manager", "Manager", "Developer", "Developer"), Ford = c(T, F, F, T), Renault = c(F, T, T, F), Ferrari = c(T, F, F, F), iPhone = c(T, T, T, F), Android = c(T, T, F, T))

#   Profession  Ford Renault Ferrari iPhone Android
# 1    Manager  TRUE   FALSE    TRUE   TRUE    TRUE
# 2    Manager FALSE    TRUE   FALSE   TRUE    TRUE
# 3  Developer FALSE    TRUE   FALSE   TRUE   FALSE
# 4  Developer  TRUE   FALSE   FALSE  FALSE    TRUE

I'd like to obtain a contingency table with the count of Car and Phone types by Profession. Of course, I am not interested in all the FALSE (or NA) values. Ideally, I'd like to present it in a table with hierarchical structure of variables, such as this:

          Manager Developer (Total)
Car
- Ford          1         1       2
- Renault       1         1       2
- Ferrari       1         0       1
Phone
- iPhone        2         1       3
- Android       2         1       3

I have tried to mess around with table but I must confess I am quite lost and don't know where to begin.

Upvotes: 2

Views: 2234

Answers (2)

user20650
user20650

Reputation: 25854

You can also do this with the reshape2 package. 

recast(dat, variable ~ Profession, id.var = 1, fun=sum, margins="Profession")

#   variable Developer Manager (all)
# 1     Ford         1       1     2
# 2  Renault         1       1     2
# 3  Ferrari         0       1     1
# 4   iPhone         1       2     3
# 5  Android         1       2     3

recast does this in one step, but to see why the variable names are in the formula have a look at 

melt(dat, 1)

and then

dcast(melt(dat, 1), variable ~ Profession, value.var='value', fun=sum)

Upvotes: 3

Vandenman
Vandenman

Reputation: 3176

This should work:

# split the data by profession, result is a list with a dataframe for every profession
data2 = split(data[, -1], data$Profession)
# colSums is then equal to the frequencies per Ford, Renault, etc.
# that is binded into a dataframe for convenience
tb = data.frame(lapply(data2, colSums))
# add a column for total
tb$Total = rowSums(tb)

Upvotes: 2

Related Questions