Re: [Boost-bugs] [Boost C++ Libraries] #2651: Documentation errors, potential application problems

Subject: Re: [Boost-bugs] [Boost C++ Libraries] #2651: Documentation errors, potential application problems
From: Boost C++ Libraries (noreply_at_[hidden])
Date: 2009-01-23 11:46:51


#2651: Documentation errors, potential application problems
------------------------------------------------------+---------------------
  Reporter: georg.mainik_at_[hidden] | Owner: johnmaddock
      Type: Bugs | Status: assigned
 Milestone: | Component: math
   Version: Boost 1.35.0 | Severity: Optimization
Resolution: | Keywords: Pareto Distribution
------------------------------------------------------+---------------------

Comment(by johnmaddock):

 Just to clarify this is an issue with getting the docs wording correct,
 not with the implementation.

 * Asuming the p values passed to the quantile function are exact binary
 values then there is no issue with accuracy, but...
 * As with all distributions if you pass a p value close to 1, and that
 value is inexact, then what we're really need in order to calculate the
 quantile is 1-p which will be mostly garbage digits for p ~ 1.

 To put this another way:

 quantile(my_pareto, 0.99)

 is potentially less accurate than

 quantile(complement(my_pareto, 0.01))

 since 1 - 0.99 != 0.01 when using fp-arithmetic.

 This assumes that the 0.01 value (for example) is either a constant, or
 computed by some manner other than subtracting 0.99 from 1.

 Not sure at present, whether we should reword this, or just remove the
 accuracy caveat from the docs, given that this is an issue with all
 distributions, and all implementations thereof not just ours.

 HTH, John.

-- 
Ticket URL: <https://svn.boost.org/trac/boost/ticket/2651#comment:3>
Boost C++ Libraries <http://www.boost.org/>
Boost provides free peer-reviewed portable C++ source libraries.

This archive was generated by hypermail 2.1.7 : 2017-02-16 18:49:59 UTC