Re: gammainv implementation (Morten Welinder)

["David J. Braden" <dbraden rochester rr com>]

Morten,
It's worse than you might realize; neither gammainv nor gammadist work 
for 0 < alpha < 1; that's not good, at all.

Further, beta is just a scale parameter; gammainv(.5,1,.2) *should* 
return 0.2*gammainv(0.5,1,1), not #NUM!.

So: we need to fix gammainv's behavior for 0 < beta < 1, and implement 
the gammadist and invgamma to work at least for 0 < alpha < 1, if not 
for all alpha except the negative integers.

TIA
Dave Braden