[R-sig-dyn-mod] Question about the NPZD model

[Faelens Ruben]

Hello Zaid,

> [...] Why do you run the model twice? Is this a requirement of the deSolve package? [...]

More often than not, you want to show the effect of a modification to the standard state of a system. For example, you want to compare the insertion of a new plant in an eco-system.

Therefore, modeling often includes two steps. First, you calculate the steady-state of an unmodified system based on an estimate of Initial Values. This initial run validates your model. It allows you to pass standard tests as:
* Does my model reproduce behavior as observed in real life?
* Does my model go towards a steady-state or recurring cyclic behavior? Does it not explode to Infinity?
* Does my model make valid assumptions about the Initial Values? Does a minor error on my Initial Values create a big difference in steady-state ? Do my initial values return in the cyclic behavior of the model?

You then use the steady-state as the IV for a modified system. This will allow you to observe the effect of your modification on the system on short-term and long-term (i.e. modified steady-state behavior).

In summary:
* IV <- RoughEstimate; SteadyState <- realSystem(IV, c(356))
* Validate minor errors on your IV do not modify the SteadyState. 
* InterestingObservations <- modifiedSystem(SteadyState, c(0,1,2,3, ...) )

This is not a requirement of the deSolve package. If you have valid and well estimated initial values for a system, you should use them. The approach above is just considered best practice when doing comparative analyses (often the case in pharmaco-economics).

IANABS: I Am Not A Bio-informatics Specialist. I am just an IT consultant working for the real specialists.