RE: [Reliable Computing] Re: Intervals "default" for beginners?
- To: "Ralph" <kelsey@xxxxxxxx>, <Jean-Pierre.Merlet@xxxxxxxxxxxxxxx>, <rbk@xxxxxxxxxxxxx>
- Subject: RE: [Reliable Computing] Re: Intervals "default" for beginners?
- From: "Kreinovich, Vladik" <vladik@xxxxxxxx>
- Date: Sat, 25 Apr 2009 20:53:15 -0600
- Cc: "Sylvain Pion" <Sylvain.Pion@xxxxxxxxxxxxxxx>, "Corliss, George" <george.corliss@xxxxxxxxxxxxx>, "Dan Zuras Intervals" <intervals08@xxxxxxxxxxxxxx>, "Alan Eliasen" <eliasen@xxxxxxxxxxxxxx>, <reliable_computing@xxxxxxxxxxxxxxxxxxxxxx>, "stds-1788" <stds-1788@xxxxxxxxxxxxxxxxx>
- Delivered-to: mhonarc@xxxxxxxxxxxxxxxx
- In-reply-to: <06D5D66D4E670ED2620F4A2B@xxxxxxxxxxxxxxxxxx>
- List-help: <http://listserv.ieee.org/cgi-bin/wa?LIST=STDS-1788>, <mailto:LISTSERV@LISTSERV.IEEE.ORG?body=INFO%20STDS-1788>
- List-owner: <mailto:STDS-1788-request@LISTSERV.IEEE.ORG>
- List-subscribe: <mailto:STDS-1788-subscribe-request@LISTSERV.IEEE.ORG>
- List-unsubscribe: <mailto:STDS-1788-unsubscribe-request@LISTSERV.IEEE.ORG>
- References: <C609F57F.74C1%George.Corliss@xxxxxxxxxxxxx> <49EB3BE3.1010305@xxxxxxxxxxxxxxx> <49F1B14D.4040207@xxxxxxxxxxxxx> <49F2484A.5050408@xxxxxxxxxxxxxxx> <06D5D66D4E670ED2620F4A2B@xxxxxxxxxxxxxxxxxx>
- Sender: stds-1788@xxxxxxxx
- Thread-index: AcnFXJmdeDuyBH0wQ5W8hYs42HrH7gAvV+Hg
- Thread-topic: [Reliable Computing] Re: Intervals "default" for beginners?
You may be interested to know that there are some packages that do
exactly what you wanted: using distributions and convolutions when we
have probabilistic information in addition to using intervals when we
only know upper bounds, the most well known is Scott Ferson's RislCalc,
Software for Risk Assessment with Uncertain Numbers, including
operations with interval, probabilistic, and fuzzy uncertainties, and
their combination
http://www.ramas.com/ramas2.htm#risk
-----Original Message-----
From Ralph
As a math student in the 60's, I stumbled across RM's book. I thought it
was an interesting mathematical system. But when I put my physicist hat
on,
I observed that intervals were a poor match for what is typically known
about an imprecisely measured quantity. As soon as I got around to it, I
planned to create a much better system using distributions and
convolutions, or something, to model the propagation of uncertainty
through
calculations.
Ralph Kelsey
Ohio Univ