Are integer vectors numeric vectors in R?

Question

I have an integer vector that I expected I could treat as a numeric vector:

> class(pf$age)
[1] "integer"
> is.numeric(pf$age)
[1] TRUE

However, when I try to use it to calculate a correlation, I get an error:

> cor.test(x = "age", y = "friend_count", data = pf)
Error in cor.test.default(x = "age", y = "friend_count", data = pf) : 
  'x' must be a numeric vector

None of my best guesses at alternate syntax work either: http://pastie.org/9595290

What's going on?

Edit:

The following syntax works:

> x = pf$age
> y = pf$friend_count
> cor.test(x, y, data = pf, method="pearson", alternative="greater")

However, I don't understand why I can't specify x and y in the function (as you can with other R functions like ggplot ). What is the difference between ggplot and cor.test ?

Answer 1

You don't refer to variables using character strings like that in a function call. You want to pass to the x and y arguments numeric vectors. You passed length 1 character vectors:

> is.numeric("age")
[1] FALSE
> is.character("age")
[1] TRUE

Hence you were asking cor.test() to compute the correlation between the strings "age" and "friend_count" .

You also mixed up the formula method of cor.test() with the default one. You supply a formula and a data object or you supply arguments x and y . You can't mix and match.

Two solutions are:

1. with(pdf, cor.test(x = age, y = friend_count))
2. cor.test( ~ age + friend_count, data = pf)

The first uses the default method, but we allow ourselves to refer to the variables in pf directly by using with() . The second uses the formula method.

As to your question in the title; yes, integer vectors are considered numeric in R:

> int <- c(1L, 2L)
> is.integer(int)
[1] TRUE
> is.numeric(int)
[1] TRUE

Do note @Joshua Ulrich's point in the comment below . Technically integers are slightly different to numerics in R as Joshua shows. However this difference need not concern users most of the time as R can convert/use these as needed. It does matter in some places, such as .C() calls for example.

Answer 2

You can use 'get' with strings to get data:

age = pf$age
friend_count = pf$friend_count

or:

attach(pf)

then following should work:

cor.test(x = get("age"), y = get("friend_count"))