Re: [Gnuplot-info] Factorial issue

[Philipp K. J. <ja...@ie...>]

But it is still meaningless, because you are
dividing by an almost equally large number
to obtain a nuber 155 orders of magnitude (!)
smaller. Round-off error will kill you.

On Monday 08 February 2010 08:18:25 am jody wrote:
> Hi
> Unless some gnuplot versions work with 32bit floats,
> I'm surprised that it wouldn't work, because 100! = 9.33262154439442e+157
> which is well inside the range of a (64bit) double...
>
> Jody
>
> >> > exponentiate at the end (if needed).
>
> On Mon, Feb 8, 2010 at 5:04 PM, Philipp K. Janert <ja...@ie...> wrote:
> > On Monday 08 February 2010 07:53:24 am Tim Wojtulewicz wrote:
> >> On Mon, Feb 08, 2010 at 07:15:21AM -0800, Philipp K. Janert wrote:
> >> > On Sunday 07 February 2010 11:07:07 am Tim Wojtulewicz wrote:
> >> > > I'm working with a fairly complicated function, one part of which is
> >> > > the division of two factorials.  I'm having an issue with though,
> >> > > because gnuplot keeps returning NaN for what should be a simple
> >> > > division.  It will return NaN for something as simple as 100!/99!,
> >> > > which should just be 100. Is there a reason for this, or an easy way
> >> > > to fix it?  Thanks.
> >> >
> >> > The intermediate results overflow, despite the fact that
> >> > the final result is of meaningful size.
> >> >
> >> > Either work out the cancellations analytically, or work with
> >> > logarithms of all quantities (Stirlings approximation!), and
> >> > exponentiate at the end (if needed).
> >>
> >> So is this a version issue then?  The reply from Thomas seems like his
> >> is working correctly.  I'm working with version 4.2.5 on OS X 10.6.
> >> It's odd that his would work, and yet mine won't.
> >
> > If anything, this is a platform issue. Quite frankly,
> > I was surprised that it worked for anybody. (It doesn't
> > for me.)
> >
> > But this is besides the point: if you need factorials
> > of large numbers (>10), then you need to find ways
> > to tame overflow issues. Unless you are working in
> > infinite precision integer arithmetic, which is not
> > something that gnuplot does.
> >
> >> Tim
> >>
> >> ------------------------------------------------------------------------
> >>--- --- The Planet: dedicated and managed hosting, cloud storage,
> >> colocation Stay online with enterprise data centers and the best network
> >> in the business Choose flexible plans and management services without
> >> long-term contracts Personal 24x7 support from experience hosting pros
> >> just a phone call away. http://p.sf.net/sfu/theplanet-com
> >> _______________________________________________
> >> gnuplot-info mailing list
> >> gnu...@li...
> >> https://lists.sourceforge.net/lists/listinfo/gnuplot-info
> >
> > -------------------------------------------------------------------------
> >----- The Planet: dedicated and managed hosting, cloud storage, colocation
> > Stay online with enterprise data centers and the best network in the
> > business Choose flexible plans and management services without long-term
> > contracts Personal 24x7 support from experience hosting pros just a phone
> > call away. http://p.sf.net/sfu/theplanet-com
> > _______________________________________________
> > gnuplot-info mailing list
> > gnu...@li...
> > https://lists.sourceforge.net/lists/listinfo/gnuplot-info
>
> ---------------------------------------------------------------------------
>--- The Planet: dedicated and managed hosting, cloud storage, colocation
> Stay online with enterprise data centers and the best network in the
> business Choose flexible plans and management services without long-term
> contracts Personal 24x7 support from experience hosting pros just a phone
> call away. http://p.sf.net/sfu/theplanet-com
> _______________________________________________
> gnuplot-info mailing list
> gnu...@li...
> https://lists.sourceforge.net/lists/listinfo/gnuplot-info