Re: More Questions about analysing ABMs

[Darren Schreiber]

The method I and my colleagues are using is to start with some set of
parameters we think is reasonable.  Then we take each of those parameters
and tweak it to an unreasonably low value and an unreasonably high value
(ceteris parabis), and a hundred values in between.  We graph all of the
output values on the y, and the range of the tweaked parameter on the x.
We compare the outputs with empirical data and also look for nonsensical
results.  We also note the dynamics of the outputs across the range of the
input.  We use these results to redefine the range of each of the input
parameters.

The next step is to explore the topology of the input parameter
interactions.  We are moving onto that step next.  We may choose to vary
pairs of the inputs, to see general ranges for interactions.  Finally, we
will do a multidimensional sweep of each of all the variables.  With just
five settings for each of the 5 input parameters, we will have 3,125
possibilities, so building our intuitions as much as possible up to this
point is crucial.  Our hope is that by the time we have swept the space we
will have a really good idea about what the model is doing and what makes a
difference.

        Darren



>address@hidden wrote:
>>
>> Steve,
>>
>> Thanks so much for your comments.
>
>A trick you might consider is to run your model over a range of values
>for some input (e.g., the magnitude of the intervention, or the severity
>of the injury) and see if you get a significant trend in results. In
>general, some of us think that trends are more useful and meaningful
>than tests for statistical significance* when looking at typically
>complex relations between stressors and responses in messy systems.
>
>*Suter, G. W. (1996). Abuse of hypothesis testing statistics in
>ecological risk assessment. Human and Ecological Risk Assessment 2(2):
>331-347.
>
>Steve
>--
>address@hidden
>Lang, Railsback & Assoc.
>250 California Ave., Arcata CA 95521
>707-822-0453; Fax 822-1868
>
>
>                  ==================================
>   Swarm-Modelling is for discussion of Simulation and Modelling techniques
>   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
>   please send a message to <address@hidden> with "help" in the
>   body of the message.
>                  ==================================


_____________________________________________

                 Darren Schreiber
                  Attorney at Law
                 Graduate Student
             Political Science, UCLA
                address@hidden
        http://www.bol.ucla.edu/~dschreib


                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================