From octave-sources-request at bevo dot che dot wisc dot edu  Thu Apr  9 13:03:09 1998
Subject: Re: sanity test for gammai 
From: Jim Van Zandt <jrv at mbunix dot mitre dot org>
To: Ted dot Harding at nessie dot mcc dot ac dot uk
cc: jrv at vanzandt dot mv dot com, octave-sources@bevo.che.wisc.edu
Date: Thu, 09 Apr 1998 14:01:42 -0400

In message <XFMail dot 980409152550 dot Ted dot Harding at nessie dot mcc dot ac dot uk>, Ted Harding writ
es:
> ...in 2.0.11 the computed value of gammai(x,a) is simply the integral
> of (t^(a-1))*exp(-t), not (as defined in A&S) of
> (t^(a-1))*exp(-t)/gamma(a) and also contrary to what "help gammai"
> says...
> 
> One can live with either interpretation of gammai, but it should be made
> definite which it is supposed to be.
> 
...
> Octave-2.0.11 gives
> 
> octave:130>  gammai(6,100)
> ans = 120.00   [ = 5! ].
> 
> 
> In octave-1.1.1 I get
> 
> octave:6>  gammai(6,100)
> ans = 1
> 
> so there has been a change along the line

Actually I prefer the new definition of gammai(a,x), which corresponds
to \gamma(a,x) in A&S (whereas the old definition corresponded to
P(a,x) in A&S).  I agree that we could live with either.

However, there's another issue.  I believe gammainc now differs from
the function with that name in MatLab.  Can anyone confirm this?  If
so, I think it's a bug.

                                 - Jim Van Zandt