On Fri, 2007-09-02 at 13:30 +0100, Stefan Nikolaus wrote:
> On Friday 09 February 2007 12:56:09 Stefan Nikolaus(UNTRUSTED, sender is
> <srs0=vgmu=hr=lists.oasis-open.org=office-formula-return-280-stefan.nikolaus=>)
> wrote:
> > gauss() is the method returning the value for GAUSS (what a surprise ;-) ).
> > And that is defined as NORMDIST(x)-0.5. NORMDIST(x) returns the value of
> > the normal distribution pdf (!) at position x. Is that what you call the
> > p-value? I think not. You want the value of the cdf! Otherwise, it makes no
> > sense at all (AFAICT).
>
> Whoops. I got this wrong: the default of NORMDIST() is to return the cdf. That
> means
> PushDouble(0.5 - gauss((mue-x)/sqrt(sigma/rValCount)));
> could be written as
> PushDouble(1.0 - NORMDIST((mue-x)/sqrt(sigma/rValCount)));
> , which is the one-tailed test and which fails for a negative argument of
> NORMDIST().
Which fails? I don't think so. Calc and Gnumeric (and according to
enough comments on the web re Excel, also Excel) calculate the
right-tail. So the p-value can be larger than 0.5.
Andreas
PS: Gnumeric currently does not implement the optional third argument
but that is an omission to be fixed.
--
Prof. Dr. Andreas J. Guelzow
Dept. of Mathematical & Computing Sciences
Concordia University College of Alberta
