Re: How to minimize the absolute sum in curve fitting?
From: Kedi (kd20128_at_yahoo.com)
Date: 07/15/04
- Next message: James Dolan: "Re: Sin Cos Tan, why not Sin Sec Tan?"
- Previous message: hack: "Re: Fixpoint for LZH"
- In reply to: Joel Bondurant: "Re: How to minimize the absolute sum in curve fitting?"
- Messages sorted by: [ date ] [ thread ]
Date: 15 Jul 2004 15:40:28 -0700
Thank you, everyone!
I cannot use any software package or human judgement for this curve
fitting, since I need to put this into a C routine to run on a device.
Many thanks to those who recommended "Iteratively Reweighted Least
Squares" and "Least Median Squares". Joel's article is very very
helpful too.
By the way, is there really no closed form solution for the "Least
Absolute Deviation" method?
I have improved the standard deviation of the estimated mean by 20% by
simply stripping some outliers. (This curve fitting is done many times
over time in the device, and I need the results to be as consistent as
possible wihtout sacrificing accuracy.) Next I am going to try
"weighted least squares" first because it is quite easy to implement.
My main aim is to reduce the standard deviation of the estimated mean
by 90%. (I don't have a good model about the noise in the data, so I
am going to test a bunch of algorithms.) I will post here again if I
achieve that.
Best Regards.
K
- Next message: James Dolan: "Re: Sin Cos Tan, why not Sin Sec Tan?"
- Previous message: hack: "Re: Fixpoint for LZH"
- In reply to: Joel Bondurant: "Re: How to minimize the absolute sum in curve fitting?"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
