On 18/12/2010 2:34 PM, casperyc wrote:>
> Hi all,
>
> #####################################
> integ=function(f,n){
> # computes vector y = f(x)
> x=runif(10000)
> y=f
> hatI=mean(y)
> hatI
> }
> # example of use
> integ(x^2+x+100,100)
> #####################################
> it returns an error says no obj 'x'
>
> how do I 'tell' R to treat 'f' as an input expression?
In integ, you can get the unevaluated expression using
expr <- substitute(f)
You can then evaluate it in the local context using
eval(expr)
So your integ function should be
integ=function(f,n){
# computes vector y = f(x)
x=runif(10000)
y=eval( substitute(f) )
hatI=mean(y)
hatI
}
I can't help saying that this is bad style, though. Using non-standard
evaluation is usually a bad idea. (There are examples like curve() in
the base packages, but they are often criticized.)
A user should be able to expect that the "x" in
integ(x^2+x+100,100)
refers to his own local variable named x, it shouldn't be a magic name.
Much better style is to require that the first argument is a function
that takes a single argument; then you'd write your integ as
integ=function(f,n){
# computes vector y = f(x)
x=runif(10000)
y=f(x)
hatI=mean(y)
hatI
}
and call it as
integ(function(x) x^2+x+100,100)
Doing this will be a lot more flexible and R-like in the long run. For
example, if you have two functions, you can say
f <- function(x) x^2+x+100
g <- function(x) x^2+x-100
and then do integ(f, 100); integ(g, 100). The code I gave you would not
work if f and g were stored as expressions:
> f <- expression(x^2+x+100)
> integ(f, 100)
[1] NA
Warning message:
In mean.default(y) : argument is not numeric or logical: returning NA
Duncan Murdoch