I have an db that look like this:
ID Group Drink
1 A yes
2 A no
3 A NA
4 B no
5 B no
6 B yes
and I would like measure how many people of group A drinks and how many people in group B drinks.
I am using length(), but this function returns 3 (NA is being considered = yes). How can I fix it?
table() is one option:
db <- read.table(text = "ID Group Drink
1 A yes
2 A no
3 A NA
4 B no
5 B no
6 B yes", header = TRUE)
with(db, table(Drink))
with(db, table(Group, Drink))
> with(db, table(Drink))
Drink
no yes
3 2
> with(db, table(Group, Drink))
Drink
Group no yes
A 1 1
B 2 1
Including the NA as a class is achieved by the useNA argument:
with(db, table(Drink, useNA = "ifany"))
> with(db, table(Drink, useNA = "ifany"))
Drink
no yes <NA>
3 2 1
You can of course store the objects returned by table() and access them as any other matrix/array:
tab <- with(db, table(Drink, useNA = "ifany"))
tab[1]
tab2 <- with(db, table(Group, Drink, useNA = "ifany"))
tab2[,1]
tab2[1,]
> tab <- with(db, table(Drink, useNA = "ifany"))
> tab[1]
no
3
> tab <- with(db, table(Drink, useNA = "ifany"))
> tab[1]
no
3
> tab2 <- with(db, table(Group, Drink, useNA = "ifany"))
> tab2[,1]
A B
1 2
> tab2[1,]
no yes <NA>
1 1 1
Here's another way using aggregate(...)
aggregate(Drink~Group,df,function(x)sum(x=="yes"))
# Group Drink
# 1 A 1
# 2 B 1
To get the percent that drink:
aggregate(Drink~Group,df,function(x)sum(x=="yes")/length(!is.na(x)))
# Group Drink
# 1 A 0.5000000
# 2 B 0.3333333
xtabs is another option:
xtabs(~ Group + Drink, df)
# Drink
#Group no yes
# A 1 1
# B 2 1
And in case you need a data.frame as output:
as.data.frame(xtabs(~ Group + Drink, df))
# Group Drink Freq
#1 A no 1
#2 B no 2
#3 A yes 1
#4 B yes 1
Assuming d is your data, and assuming NA is considered yes (since you stated that in your post), the proportions of drinkers are
> d$Drink[is.na(d$Drink)] <- 'yes'
> tab <- table(d$Group, d$Drink)
> tab[,'yes']/rowSums(tab)
## A B
## 0.6666667 0.3333333
You could also play around with the count function in package plyr
> library(plyr)
> x <- count(d)
> cbind(x[x$Drink == 'yes', ], inGroup = count(d$Group)$freq)
# Group Drink freq inGroup
# 2 A yes 2 3
# 4 B yes 1 3
You could also make use of the function prop.table() which will add proportions to your table values that are returned.
