recap on Rosen

Posted by Marcus G. Daniels on Apr 26, 2008; 4:36am
URL: http://friam.383.s1.nabble.com/Welcome-Jim-tp526087p526121.html

phil henshaw wrote:

> Glen wrote:
>  
>> I believe so.  At least 1/2 of the solution to any problem lies in a
>> good formulation of the problem.  And in that sense, being able to
>> state
>> (as precisely as possible) which closures are maintained in which
>> context and which closures are broken in which context, therefore,
>> contributes immensely to the solution.
>>    
>
> [ph] the requirement is that your model describe new behavior of independent
> organisms or communities things you have no information about because they
> never occurred before.  What's the modeling strategy for that?
>  

Find a function that well describes a state of a thing or aggregate
measurement of interest at t - 2 that gives the state at t - 1 that
gives a state at t.  Then prediction is a matter of applying the
function more times.  Add more functions to describe more individual
things or aggregates and note when there are shared functions in those
definitions (e.g. food web fundamentally depends photosynthesis).

If you want to define all things to be independent, then there is no
point in talking about interactions -- you've already defined away the
possibility of that!    Covariance is zero.

Marcus