QUERY Measurement, Calibration and Validation (SD6124)

[Monte Kietpawpan kietpawpan yahoo.com]

Posted by  Monte Kietpawpan <kietpawpan at yahoo.com>
   
  Dear members:
   
  I have only limited experience in system dynamics modeling and am 
  interested in views to confirm or to correct the below note, in which
  I wish to build confidence.
   
  "Model calibration is hard to perform when the situation being modeled
  is a rare event, when empirical data cannot be collected at the time of 
  the modeling, or when the parameter values are highly uncertain due to
  the dynamics of the system structure. For example, the values of walking speed of 
  pedestrians in a tourist desination may vary when different population
  of tourists occupy the area over the year. I mean we cannot collect samples that 
  represent the population in the area because there are various populations to occupy
  the same area at different periods of time. 
   
  The calibration (reproductivity test) will be necessary when
  the structural validity test is hard to conduct because there is no accepted 
  physical laws or equations that govern the dynamics of concern. Here behavioral
  validity test is perhaps a practical way to build confidence in the model.
   
  When individual governing equations have been accepted, we just integrate 
  them into a simulation model and need not to calibrate the model results through 
  a full-scale field experiment, or the model will be less useful if a major 
  purpose of the model is to represent a complex structure so that we will not 
  have to conduct a full-scale field experiment."
   
  I hope somebody will help discuss on this issue.
   
  Monte Kietpawpan
   


Monte Kietpawpan
PSU, Thailand
Posted by  Monte Kietpawpan <kietpawpan at yahoo.com>
posting date  Mon, 1 Jan 2007 21:31:47 -0800 (PST)