p.1 Modeling is neither science nor mathematics; it is the craft that builds bridges between the two...
The Art of Modeling consists of matching the behavior of computational processes to that exhibited by series
of measurements.
p.2 Analyzing complicated systems need not require great mathematical sophistication or intellectual
brilliance. But it does require clear thinking, the kind of thinking that distinguishes sharply between the dynamics
of change and the unchanging dynamical principles needed to analyze change.
p.3 In many cases there are no scientific "laws" from which to deduce equations, or these "laws" are too
vague to be more than general constraints. The inductive problem is, "Given the solution, what are the equations?" This is
what science is all about... The modeler has to invent some equations to match the data... The first thing
to do is establish an error budget. Some kind of criterion needs to be set to say, "This is signal, but this other thing is
noise." ... A first caveat is not to have the expectations of doing prediction as accurately as celestial mechanics or some
of the other physical sciences... Model building cannot begin without the participation of the human mind. And the
mind cannot comprehend a system of everything affecting everything else.
p.4 After setting a rather high error budget, the modeler should attempt to find a few cause-and-effect
relationships. Then start looking for some feedbacks. Build up a model bit by bit; don't try to have it fully
mature at creation, like the goddess Aphrodite. In many cases it will be a good idea to look at short time spans
for "cause and effect." Then try to spot the places where the feedbacks become significant.
p.4 A lot of thought should be devoted to the uses which the model will be put. What decisions will be made
depending on the parameters or forecasts coming from the model?
p.4 Knowing something about the subject matter you are attempting to model usually is helpful. But sometimes
knowing too much is not. A fresh perspective always is useful. Look into what other disciplines are
doing. Often they have the same problems you do, and they may have some ideas you can use.
p.4 In many cases the quickest thing to do is cobble together something that works, however inefficiently,
and then replace parts of it with better techniques as you find them.
p.5 Always remember that a model is not reality, but something that imitates reality at a certain
scale.
p.59 Intuition and judgment play an important role at every step of the modeling process
p.74 Complex dynamic systems will require the combination of deterministic models with these statistical
ones to produce the best feasible interpolations and forecasts.
p.81 A stable core model is constructed that predicts things in the aggregate. The deviations
from this are treated as noise-driven, damped processes from the physical point of view.
p.352 Forecasts serve many purposes. The most important is to predict the future state of something
as well as possible, using existing knowledge. This is a primary goal of modeling.
