[Ns-bugs] [Bug 181] Normal Variable Infinite Value & Bounds

[bugzilla-daemon@nsnam-www.ece.gatech.edu]

http://www.nsnam.org/bugzilla/show_bug.cgi?id=181





------- Comment #6 from mathieu.lacage at sophia.inria.fr  2008-06-10 12:38 -------
(In reply to comment #5)
> Created an attachment (id=157)
 --> (http://www.nsnam.org/bugzilla/attachment.cgi?id=157&action=view) [edit]
> test case for above patch
> 
> This is a test case which generates both unbounded and bounded versions of a
> Normal(10,2).  Histogram plots verify that it works at bounding the
> distribution symmetrically about the mean.  The tail of the bounded
> distribution is skewed however:
> http://www.prism.gatech.edu/~gtg037s/files/normal-variable.eps
> 
> This is due to the way the code is forcing values outside of the range to be on
> the range boundary.  Ideally, the code would drop values outside of the range
> and "try again" so that the boundary didn't have this behavrior.
> 
> Honestly though, this API was never intended to do this.  I conceived that the
> bound would be set somewhere like 5 standard deviations away, so that this tail
> effect is minimized.  What do people think?

>From the perspective of the user, I think that forcing values outside of the
range to be on the edge of the range is actively harmful. Can't you just drop
values outside of the range and retry (from within the random variable code)
and return a value when it is within the range only ?


-- 
Configure bugmail: http://www.nsnam.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.