home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky sci.math.stat:2350 sci.math:14879
- Newsgroups: sci.math.stat,sci.math
- Path: sparky!uunet!utcsri!helios.physics.utoronto.ca!alchemy.chem.utoronto.ca!mroussel
- From: mroussel@alchemy.chem.utoronto.ca (Marc Roussel)
- Subject: Re: Repost: Algorithms for model discrimination
- Message-ID: <1992Nov13.012349.15438@alchemy.chem.utoronto.ca>
- Followup-To: sci.math
- Organization: Department of Chemistry, University of Toronto
- References: <BxKowz.7Ho@constellation.ecn.uoknor.edu> <BxLHMG.K9J@constellation.ecn.uoknor.edu>
- Date: Fri, 13 Nov 1992 01:23:49 GMT
- Lines: 30
-
- In article <BxLHMG.K9J@constellation.ecn.uoknor.edu> u235@rex.chb.uokhsc.edu
- (Yung-Jin Lee) writes:
- >I have several rival models which are needed to to
- >be discriminated in my kinetic studies. Is there any numerical
- >algorithms (Fortran in preferred) can be used in my studies? What
- >I have tried is the method of sequential experimental design from
- >a text of 'Mathematical Modelling and Computers in Endocrinology'
- >written by McIntosh JEA and McIntosh RP, 1980, Springer-Verlag.
-
- I don't have an algorithm to offer, but there is an analytic method
- which will sometimes work. Here is the reference:
-
- M.R. Roussel and S.J. Fraser, "Accurate steady-state approximations:
- Implications for kinetics experiments and mechanism", J. Phys. Chem. 95,
- 8762-8770 (1991).
-
- The heart of the method is a recasting of the ODE problem into a
- functional equation problem. When it works, it works spectacularly.
- When it doesn't work, you either have a system with unusual dynamics
- (and that is valuable to know as well) or the business of distinguishing
- the two mechanisms will be very difficult by any method.
- If you have the right equipment, you can also try fast relaxation
- methods. Have a look at K. Hiromi, Kinetics of Fast Enzyme Reactions
- (Wiley: New York, 1979), esp. pp. 41-42.
-
- Marc R. Roussel
- mroussel@alchemy.chem.utoronto.ca
-
- P.S.: I have redirected followups to sci.math since this has little to
- do with statistics.
-