Re: [Boost-bugs] [Boost C++ Libraries] #6992: accumulator's median feature skips 1st two data points.

Subject: Re: [Boost-bugs] [Boost C++ Libraries] #6992: accumulator's median feature skips 1st two data points.
From: Boost C++ Libraries (noreply_at_[hidden])
Date: 2012-11-15 20:43:42


#6992: accumulator's median feature skips 1st two data points.
----------------------------------+-----------------------------------------
  Reporter: polyactis@… | Owner: troyer
      Type: Bugs | Status: new
 Milestone: To Be Determined | Component: accumulator
   Version: Boost 1.41.0 | Severity: Problem
Resolution: | Keywords:
----------------------------------+-----------------------------------------

Comment (by troyer):

 Eric, having dinner with you in Redmond is dangerous especially if I agree
 to take a look at the open accumulator tickets. Here is my conclusion
 after looking at the issue:

 First, median estimated are notoriously hard and you never get an exact
 median unless you store at least half of the samples. Hence, unlike the
 mean which can easily and unambiguously be estimated (as long as the
 variance is finite), median estimation is harder and different algorithms
 to estimate the median will give different results.

 Our default estimator is a P^2 quantile estimator, which only stores and
 updates five numbers and hence has a minimal memory footprint. However, it
 requires at least five samples before it gives sensible output and I am
 thus not surprised that using less than five samples does not work.

 Shall we throw an exception if the count is less than 5, or just document
 it more clearly?

-- 
Ticket URL: <https://svn.boost.org/trac/boost/ticket/6992#comment:2>
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 18:50:11 UTC