Author 
Comment/Response 
Chris Reiter

01/26/99 8:43pm
As I understand the ''grammar'' of the symbol NonlinearFit as described in usage for the package of the same name, data must be in the form of a vector or a matrix. If in the form of a vector, the stated data {f1,f2,...,fn} is the dependant response, f, with independent control implied as 1,2,3,...,n. If a matrix, (and here is the problem) the dependant data (x,y,...,z) is explisit {{x1,y1,...z1,f1},...}} and the response is the last stated,f. These rules imply NonlinearFit can only fit one data set at a time, i.e., only one graph at a time. I wish to fit several data sets simultaneously (one merit function minimisation, several data sets) such that some, but not all, of the parameters are shared (slaved across data sets). For example, I could start a particular chemical reaction and follow it's progress spectrophotometriclly at multiple wavelengths. I could then perform an SVD on this rectangular matrix to give me several time vectors made up of a linear combination of the reaction processes. I would then have to fit a model of how I think the reaction is going to all of these time vectors in one shot. This is because some ''amount'' of the reaction could be in one basis vector and the rest in another. Thus to get one optimization for one parameter, where the parameter is in several data sets, I have to do a simultaneous fit. (I can do simultaneous fitting on multiple data sets with parameter sharing in a program called Origin, but the version I have does not give me singular value decomposition, and I have not done much MathLink experimentation.) Have I made any sense? I have a feeling that I am overlooking some simple script such as putting my model in matrix form that would fix my problem.
Thank you for your time.
Chris Reiter
PS could you email me a responce as well as post?
URL: , 
