
Advanced Statistics Advanced Probability and Statistics Math Forum 
 LinkBack  Thread Tools  Display Modes 
January 24th, 2012, 04:22 AM  #1 
Newbie Joined: Jan 2012 Posts: 2 Thanks: 0  Goodness of fit metrics for comparing nonlinear data series
Apologies if this is an inappropriate place to post my question. I have ~650 concentration vs time profiles where dt increases in an irregular manner with t. The values of a subset of the parameters that I believe dictate the shape of this profile are known. A model of the underlying process exists. This model can be supplied with known parameter values and parameter estimates to produce simulated concentration profiles (with v. small dt) that can be compared to real profiles. Expressing this model algebraicly will not be trivial, if possible at all. It is presently evaluated in an iterative manner. The real and simulated concentration profiles look very approximately like exp(x) for x < a value and like 1exp(x) for x >= that value. I wish to find a goodness of fit metric that allows me to compare a real concentration profile to a corresponding simulated one, preferably one that is suitable for application to nonlinear data (unlike the leastsquares error) and is somewhat 'shapeaware' i.e. a 'soft' metric. Any suggestions? (The end goal is to find a method for identifying the most likely theoretical concentration profile given a sparse and noisy real profile.) Regards, Will Furnass 

Tags 
comparing, data, fit, goodness, metrics, nonlinear, series 
Thread Tools  
Display Modes  

Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Similarity of time series data  Challenger85  Real Analysis  0  September 29th, 2013 11:11 PM 
Fourier series for array of data  ElSeba  Math Software  0  February 16th, 2011 10:22 AM 
How to calculate simple linear model for noisy data?  stephanos  Advanced Statistics  4  July 22nd, 2010 05:54 AM 
Help on comparing differing sets of data  carter22  Algebra  2  March 1st, 2010 05:49 PM 
Comparing multiple data sets with different n values  cboldon  Advanced Statistics  2  April 7th, 2008 03:01 PM 