Re: Robustness Check

[Paul Johnson]

Rick Riolo wrote:
> 
> for benedikts 1-2, we provide drone and for 3 we also
> have some scripts you could start from and probably hack
> in no time to do what you want.
> for pages on how we use drone with some "experimental tool"
> classes to run multiple experiments, see
>   http://www.pscs.umich.edu/LAB/Doc/SwarmStuff/UM-swarm-index.html
> and for tools for simple data plotting and analysis, see under
>   http://www.pscs.umich.edu/LAB/doc-index.html
> the pages
>   - A short note on using Drone for doing experiments.
>   - How to do simple plots and analysis of simulation data.
> 
Drone itself is very easy to install. As I recall it involves checking
the first line of the drone script to see that it matches your system. 
It is tougher to see how to make your application work with Drone, but
it is highly feasible.

Since it seemed like I would bust getting this to work during the last
week, let me add this bit.  I did not want to wade through the
UM-specialized stuff (the parameter manager, especially) in order to
understand how to repeat a simulation.  If I could be in Rick's lab, I
would do it for sure because I could get him to explain all the details
to me (or die trying...).  I wanted a direct approach to repeat a
simulation with a minimum of fuss.  

I now have a working small example of drone you can see/try and it does
not require HeatbugsPlus or any other special libraries.  I posted a
note about how to do it last week in the swarm-support list, it had a
subject like "Drone worked: here's how".  Drone will nicely create the
random seeds that your program can absorb and it has built in features
for repeating things.

As for the data analysis part, I agree with the others that you should
do that outside swarm. I'm very frustrated at the moment at this stage
because I don't know perl or awk and I'm starting to get disgusted that
I have to learn a whole new language every time I learn around.  I
didn't start on this project to become a programmer/systems
administrator, but that's what it seems I am heading toward.  ( I keep
telling myself I can get new jobs if I keep studying.  The problem is
that some people say 'learn perl <or insert language here>, its great'
and others say 'perl <other> is an abomination on the face of the earth
and people who use it are clumsy oafs.'  

If you want to take the easy route, let me tell you what I would do.  I
expect that, in an emergency, I will do this myself. If you can use
fprintf statements to output the data into one giant file, your
situation is easily managed if you can get  your hands on any of the
popular/widely used statistical packages.  I have much experience with
SAS, SPSS, and Axum (the graphical precursor to S+4.0).  Each of these
has easy ways to take in an ascii file, and then pick and choose rows
for analysis.   As long as the data is in one monstrously large file,
there is no big problem.  When you output the numbers, you make sure
each line has markers for the simulation run and other identifiers.  My
advice here is not purely hypothetical. Three years ago I used this
approach to manage data from a simulation that I had written in turbo
pascal and I make photo ready images with Axum and the people at the
Journal of Theoretical Politics did not give me any abuse.

If the data is split across many directories, then it is tougher.  I'm
going to see if I can steal the UM scripts that Rick mentioned. I also
have some perl scripts donated by another kind member of the list, I'll
investigate to see what they do. 

I still have not done much with R, but I've been in their email list for
about 3 months and it looks like a pretty steep learning curve.  I might
ask in there if they have constructed a "data retrieval" module that can
grab a certain line or column from a data file.  Here's why I guess such
a think might exist or could be created.

Axum has a scripting language built into it that, I think, can traverse
through directories and open up datasets.  I've not done that within the
last 12 months, but I know it worked in the old days.  If you ended up
with data scattered across 100 directories, that would be the way to go.
I only thought of Axum because Axum was bought up by Mathsoft and it
supplied the graphical components of S+4.0, and S+4.0 seems to have a
scripting tool too.  But I don't have a windows machine handy with which
to try it lately. 

-- 
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

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.