BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Is my model too complex? Evaluating model formulation using model 
 reduction - Professor Neil Crout\, School of Biosciences\, University of N
 ottingham
DTSTART:20110606T100000Z
DTEND:20110606T110000Z
UID:TALK31603@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:There is wide acceptance that models which seek to represent b
 iological or environmental processes should be evaluated before they are a
 pplied. Numerous technical methods have evolved in order to address this r
 equirement. Literature in this area is further supplemented by more philos
 ophical discussions on the role of model evaluation/validation given that 
 models are (nearly always) known to be approximate at best. \nWhile mechan
 istic models tend to be detailed\, they are less detailed than the real sy
 stems they seek to describe\, so judgements are being made about the appro
 priate level of detail within the process of model development. These judg
 ements are difficult to test\, consequently it is easy for models to becom
 e over-parameterised\, potentially increasing uncertainty in predictions. 
 \nWork at Nottingham has sought to address these difficulties. We propose 
 and implement a method which explores a family of simpler (reduced) models
  obtained by replacing model variables with constants. The procedure itera
 tively searches the simpler model formulations and compares models in term
 s of their ability to predict observed data. Under appropriate assumptions
  the procedure can be implemented within a Bayesian framework enabling the
  results to be summarised as model probabilities and replacement probabili
 ties for individual variables which lend themselves to mechanistic interpr
 etation. This provides powerful diagnostic information to support model de
 velopment\, and can identify areas of model over-parameterisation with imp
 lications for interpretation of model results. \nThe method has been appli
 ed to a range of different example models. In each case reduced models are
  identified which outperform the original full model in terms of compariso
 ns to observations\, suggesting some over-parameterisation has occurred du
 ring model development. We argue that the proposed approach is relevant to
  anyone involved in the development or use of process based mathematical m
 odels\, especially those where understanding is encoded via empirically ba
 sed relationships. 
LOCATION:Small lecture theatre\, Microsoft Research Ltd\, 7 J J Thomson Av
 enue (Off Madingley Road)\, Cambridge
END:VEVENT
END:VCALENDAR