Other pages: Bio/Geo/CS250, MainModelingWiki
Some things to keep in mind while you think about what to propose for your individual modeling project:
-
Start by thinking of a natural or social phenomenon that interests you and about which you already know something.
-
Alternatively, start with a phenomenon about which there's something that puzzles you.
-
Whatever it is, make sure that it's a single phenomenon which is very easily recognized.
-
You want it to be easily recognized because you want to be able to tell definitively whether or not your model has successfully produced the phenomenon.
-
It might be easily recognized because it's a yes/no sort of thing -- like a species goes extinct or a material is resistant to cracks.
-
It might be easily recognized because you've come up with a statistical definition of the phenomenon -- like some population of genetic loci accumulates mutations at a significantly higher rate than blah blah blah.
-
Think small.
-
Choose a system with few different kinds of components. (Note that that's different from few components. You could easily do a model of, say, a population of millions of individuals -- as long as there are only a few different kinds of individuals to keep track of.)
-
Choose a system in which you understand pretty well how the components interact with each other.
-
And again, choose a phenomenon which you'll be able to say yes or no, the model did or didn't reproduce it.
-
People want different things from their models: mechanistic realism, quantitative accuracy, explanatory power, and generality. Usually you can accomplish one of these things, but never three or four. You'll almost certainly have to sacrifice two or three of these potential goals.
-
If you decide that you're after quantitative accuracy, pick a system that is very, very simple: few components, few interactions. You may be spending a lot of time fine-tuning the mathematical representations of the interactions (and finding the data describing them), and you won't want to do it for a large set of interactions.
-
Also, if you're after quantitative accuracy, be good at math.
At this stage, give me a paragraph describing the phenomenon you're interested in modeling -- what you know about it and what kinds of questions you have about it, if any.
