Re: [boost] More Math Functions for Statistics (and TR2)

From: Paul A Bristow (pbristow_at_[hidden])
Date: 2005-12-04 09:43:49

| -----Original Message-----
| From: boost-bounces_at_[hidden] 
| [mailto:boost-bounces_at_[hidden]] On Behalf Of Thorsten Ottosen
| Sent: 03 December 2005 18:32
| To: boost_at_[hidden]
| Subject: Re: [boost] More Math Functions for Statistics (and TR2)

| Thanks for doing this important work.

Thanks for your encouragement.

| Q: is the precision of the algorithms actually so good that 
| it makes a 
| difference if one uses float, double or long double?

The results shown in the package strongly suggest so, but some functions are
MUCH harder to get right than others.
| 
| I guess throwing is out of the question because of C compatibility?

:-((

| How are error handled in the new tr1 math functions?
C style matherr, I believe, but I have yet to see examples coded.  I suspect
Dinkumware may have built in options to use exceptions?
 
|  > my impression is that most of the warnings look spurious.
| How many warnings? what type of warning?

Dozens - mostly data loss on conversion, unsigned/signed, untidynesses like
unused parameters.  I don't want to alter Moshiers code at all (yet).

| When you write a revised paper, please enhance the "motivation" section
| by drawing from exampes in other languages. If it has been part of 
| Visual Basic, then write it. If similar functions are used often in Java, 
| then tell something about it.

OK - but there isn't much to tell except that they have them.
 
| 1. why so many functions? are they all really needed?

Well of course there was some hyperbole about the number - it counts up from
sin and cos to the highest functions - but they all link together so any one
missing is a real pain to someone.

| 2. would it make more sense to standardize a usability layer on top 

IMO, higher layers are entirely a user matter.  The incomplete beta is to
statistics as  sin and cos are to geometry.

One might consider something to get the moments - mean, variance, skew and
kurtosis - but then you get into dealing with the many containers and
iterator solutions look attractive, but then where do you put the calculated
moments?  I'd rather not go there - yet:  the functions are probably more
than I can chew ;-)

Paul