Hi,
On Mon, 2009-01-12 at 15:50 +0100, Eike Rathke wrote:
> On Sunday, 2008-12-21 16:50:05 -0700, Andreas J. Guelzow wrote:
>
> > On Sun, 2008-12-21 at 22:57 +0000, wrote:
> > > - NEGBINOMDIST()
> > > - Changed parameter names to common names x,r,p.
> >
> > I would think it would have been more important to fix the constraints:
> > How can any negative r make sense here?
>
> So far no one complained, so I wasn't aware.. it seems that both, x and
> r, shouldn't be negative.
>
I think we should definitely allow negative x. IMHO all the
distributions should be defined for all values of x: these
distributions/densities. If x is not in the support then the returned
value (in the non-cumulative case) should be 0, while in the cumulative
case it should have the obvious value.
Andreas
