How to minimize the absolute sum in curve fitting?

From: Kedi (kd20128_at_yahoo.com)
Date: 07/15/04 

Date: 14 Jul 2004 19:59:06 -0700

Hi,

I was using simple linear regression to fit some data to a straight
line on the X-Y plane, but the results are not satisfactory because of
some big outliers. It seems that trying to minimize the sum of the
absolute deviations will be a lot more useful in this case.

I vaguely remember that most of these non-linear minimization/curve
fitting problems have no closed form solutions and require iterations
to approach the answer ......

Can someone point me to an algorithm that is easy to implement? It
will be even better if that algorithm converges fast.

A big THANK YOU in advance!

Best Regards.

K

Relevant Pages

  • Re: How to minimize the absolute sum in curve fitting?
    ... > I was using simple linear regression to fit some data to a straight ... > line on the X-Y plane, but the results are not satisfactory because of ... > will be even better if that algorithm converges fast. ...
    (sci.math)
  • Re: Corridors can be hard to make
    ... I have to use some kind of algorithm to determine ... > that the corridor goes through a "straight" wall.. ...
    (rec.games.roguelike.development)
  • Re: Measuring the randomness of a signal...
    ... Nathan Baulch wrote: ... > I need an algorithm that can measure the randomness of a signal ... > straight line and 1 represents white noise. ... You should enable all the tests in RaBiGeTe, ...
    (sci.math)
  • Gears Method/BDF Algorithm
    ... I am in need of an algorithm to numerically solve a set of differential ... I programmed the 4step adaptive Runge-Kutta, ... and the results I get from mathematica using BDF ... are satisfactory. ...
    (sci.math.num-analysis)
  • Re: Combinatorics problem
    ... >and add them up, call this sum k. ... It seems pretty straight forward that you can generate every possible ... k between these two extremes by the following algorithm ...
    (sci.math)