Re: [Swarm-Modelling] Swarm Modeling

[Paul E. Johnson]

I think it will be really interesting to see your tutorial.  As my class 
is winding down now, I have  a few pressing tutorial elements to get 
ready for them.  I believe in the simpleObserverBug3 element I added in 
cvs tutorial.  And now it occurs to me I need to write 
simpleObserverBug4 (or some such) about how to write a batch observer 
swarm onto it.  

After that is where the problem starts.  How to collect data and 
present?  We have gone around here in this list sometimes on this.  I 
believe the tutorial's simpleExperBug tutorial is not useful and 
misleading in many ways.  It is very difficult to get a GUI model to 
cleanly restart.  That's why I like the simpleObserverBug3 approach. 
But after that, we need a survey of ways people run models over and 
over again and then summarize/present results for publication.  

I know it is simplistic, but I still  find it easiest to write out 
numbers into a file for each run of the simulation as described here:

http://lark.cc.ku.edu/~pauljohn/ps909/Agent-Based_Models/fprintfnote.txt

and I have used some scripts to "grab the last line" and collect them 
into a file to make summaries.  OR I have code for R to step one-by-one 
through the output datasets and show the patterns I'm looking at.

I promised the students that by Wednesday, 5pm I would cobble together 
my notes on this, I'll share what I have. When you have some stuff, 
please announce so I can see too.

There are some problems I did not solve yet.  If a model has some 
natural stopping point, one can take measures, and then summarise across 
runs.  If a model has no natural stopping point, then it is a difficult 
thing to know when to stop and measure. Also, the development of 
measures is not so easy as most students expect.  


David H. Clements wrote:

> I am putting together a tutorial that takes people through the 
> process  of model design and implementation using Swarm and I was 
> wondering what  people used as their software development methodology?
>
> The one that I have seen the most is a variant on the rapid 
> prototyping  and waterfall paradigms, which looks something like this:
>
> Concept->Requirements<->Specification<->Design<->Implementation/Rapid- 
> Prototype<->Data Collection<->Data Analysis<->Testing/Sanity Check  
> (this also connects back to requirements)->Final Analysis.
>
> Any feedback is appreciated!
>
> Sincerely,
> David


--
Paul E. Johnson                       email: address@hidden
Dept. of Political Science            http://lark.cc.ukans.edu/~pauljohn
University of Kansas                  Office: (785) 864-9086
Lawrence, Kansas 66045                FAX: (785) 864-5700