help-gsl
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Help-gsl] 1st attempt at fitting a Gaussian function with GSL.


From: James Bergstra
Subject: Re: [Help-gsl] 1st attempt at fitting a Gaussian function with GSL.
Date: Wed, 21 Dec 2005 18:12:45 -0500
User-agent: Mutt/1.4.2.1i

Hi Dave,

Maybe we could take this thread off of gsl-help?  Your issues seem more related
to statistics and C programming, and less with the GSL.

I'll be happy to help you via personal correspondance.

James

On Wed, Dec 21, 2005 at 04:22:57PM -0600, David Eric Miller wrote:
> Hi James,
> 
> You are quite correct.  I am attempting to do a Chi Squared 
> minimization
> parameter estimation.  I did not put all of the necessary pieces 
> into
> my previous email since I was just starting to figure out how to
> use gsl.
> 
> If you could take a look at my last posting, I have a more 
> complete
> test program which I am trying to get working.
> 
> Thanks again.
> 
> -dave miller
> 
> 
> On Tue, 20 Dec 2005 15:59:26 -0600, James Bergstra  
> <address@hidden> wrote:
> 
> >>> >I need to fit a Gaussian function to an array of data points.
> >>>
> >>>What is the nature of your data? Are they just X values or X,Y 
> >>pairs?
> >
> >>The data consists of (x, y) pairs with an associated error on y.
> >
> >Ok, so you are doing what is called parametric model estimation. 
> >Your  model of
> >y-given-x is a gaussian with some parameters, and you want to 
> >know what  are
> >those parameters.  The way to use the function minimizers to get 
> >those
> >parameters is to define a differentiable LOSS FUNCTION.  This 
> >loss  function will
> >say for each (x,y) of data, how bad a given parameter is.  You 
> >want to  find the
> >parameter such that the total loss over all your points is the 
> >smallest.
> >
> >The problem with your initial posting was that there was no loss 
> >function to be
> >seen, and none of the functions you wrote was the one to 
> >minimize!
> >
> >A textbook on statistical inference can give you a better sense 
> >of what  is going
> >on.  (check out Wasserman's "All of Statistics", or Bishop's 
> >book on  neural
> >networks.)
> >
> >James
> >
> 
> 
> 
> 
> _______________________________________________
> Help-gsl mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/help-gsl

-- 
james bergstra
http://www-etud.iro.umontreal.ca/~bergstrj





reply via email to

[Prev in Thread] Current Thread [Next in Thread]