[R] over/under flow

[William H. Asquith]

Great, but to followup, how do I select the bounds (B,E) on the root 
for an arbitrary machine?

OVER = uniroot(function(x) lgamma(x)-log(.Machine$double.xmax), 
c(B,E))$root

If uniroot() is fast enough, is it appropriate for me to set B at say 1 
and E at log(.Machine$double.max)?  Suggestions on do this the proper 
"R way"?  Perhaps this . . .

OVER = uniroot(function(x) lgamma(x)-log(.Machine$double.xmax), 
c(1,log(.Machine$double.xmax)))$root

I am working on a package so different machines will be involved thus 
simple 171,172 might not be the best idea for the root?

THANKS!

-wha


On Jul 3, 2005, at 3:43 PM, Peter Dalgaard wrote:

> "William H. Asquith" <wasquith at austin.rr.com> writes:
>
>> I am porting some FORTRAN to R in which an Inf triggers an if().  The
>> trigger is infinite on exp(lgamma(OVER)).  What is the canonical R
>> style of determining OVER when exp(OVER)== Inf?  The code structure
>> that I am
>> porting is best left intact--so I need to query R somehow to the value
>> of OVER that causes exp(lgamma(OVER)) to equal Inf.
>>
>> On my system,
>> exp(lgamma(171)) is about first to equal Inf.
>>
>> I asked similar question a few weeks ago on exp(OVER) and got the
>> answer back as log(.Machine$double.xmax).  I now have the lgamma
>> involved.  I think that answer is what is OVER such the
>>
>> .Machine$double.xmax = lgamma(OVER),
>
> Not quite... (see below)
>
>> but I am not sure how to invert or solve for OVER
>
>
>> uniroot(function(x) lgamma(x)-log(.Machine$double.xmax), c(171,172))
> $root
> [1] 171.6244
>
> $f.root
> [1] -1.462051e-07
>
> $iter
> [1] 3
>
> $estim.prec
> [1] 6.103516e-05
>
>
> -- 
>    O__  ---- Peter Dalgaard             Øster Farimagsgade 5, Entr.B
>   c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
>  (*) \(*) -- University of Copenhagen   Denmark          Ph: (+45) 
> 35327918
> ~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)                  FAX: (+45) 
> 35327907
>