[R] problems with integrate ... arguments

[Duncan Murdoch]

On 09/04/2009 8:07 AM, Richard Morey wrote:
> Hi everyone,
> 
> I saw this problem dealt with here:
> http://markmail.org/search/list:r-project?q=integrate#query:list%3Ar-project%20integrate+page:1+mid:qczmyzr676pgmaaw+state:results
> 
> but no one answered that request that I can tell. I'm having the same
> problem. I'm having problems passing arguments to functions that I'd
> like to integrate. I could swear this worked in the past, but I can't
> get it to work now. I'm running R on Windows.
> 
> Here's my toy code to reproduce the problem:
> 
> 
> R.Version()$version.string
> # [1] "R version 2.8.1 (2008-12-22)"
> 
> f=function(x,const)
> 2^x+const
> 
> f2=Vectorize(f,c("x"))
> 
> testval=2
> 
> f2(1:5,testval)
> # [1]  4  6 10 18 34
> # The vectorized function seems to work right.
> 
> integrate(f2,.5,1,c=2)
> # Works. Returns
> # 1.845111 with absolute error < 2.0e-14
> 
> integrate(f=f2,.5,1,const=testval)
> # Doesn't work. Returns
> # Error in eval(expr, envir, enclos) :
> #  ..1 used in an incorrect context, no ... to look in
> 
> I understand that I don't have to Vectorize this function for it to
> work, but the more complicated functions I'm using require Vectorization.
> 
> Why can't I pass the argument "const" as a variable, instead of a value?
>   I don't understand this error.

I'm not sure what's happening exactly; both Vectorize and integrate do 
tricky things with evaluation frames, and it looks as though they are 
not interacting well.

As a workaround, I'd suggest using a single argument function f instead:

 > makef <- function(const) {
+   return(function(x) 2^x + const)
+ }

makef(testval) will capture the value of testval, so we don't need to 
pass it, and then things work:

 > f <- makef(testval)
 > f2 <- Vectorize(f)
 > integrate(f=f2, 0.5, 1)
1.845111 with absolute error < 2.0e-14

Duncan Murdoch