s-news
[Top] [All Lists]

Re: Numerical Approximation for cumulative t-distribution?

To: Tom Downing <td@quaestor.com>
Subject: Re: Numerical Approximation for cumulative t-distribution?
From: ripley@stats.ox.ac.uk
Date: Tue, 25 Jun 2002 21:41:02 +0100 (BST)
Cc: <s-news@lists.biostat.wustl.edu>
In-reply-to: <63EC49A6CC83D84482C9817DCF49B072028B06@paris.quaestor.com>
On Tue, 25 Jun 2002, Tom Downing wrote:

>       Aside from numerical integration, would anyone
> know if there exist algorithms to calculate the cumulative
> t-distribution density function( implemented via the
> s-plus pt(q,df) function) ?   I am looking for something

Yes, there are several.  I would look in the (older) books on statistical
computing.  The t is a special case of the F and so can be computed from
incomplete beta functions, and you can find algorithms for those on
Abramowitz & Stegun (from memory).

There are the Applied Statistics algorithms AS3 and 27 for lower and upper
tails respectively.  I also did a search on netlib (GAMS category L5alt)
with no hits.

> akin to the relatively well-known polynomial
> approximations for the cumulative normal.  Or would it
> be quicker to actually do the integration
> with something like gaussian quadrature ?

(Not by a very large margin.)

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595


<Prev in Thread] Current Thread [Next in Thread>