Some functions which are difficult to integrate by traditional methods
can be handled by Monte Carlo methods. If you want to know the value of an
integral of y(x) between a and b, and you know that y is bounded between 0
and h, you can take a series of N (x;random,yrandom)
pairs at random from the space
x>a, x<b, y>0, y<h. Let S be the number of (xrandom,yrandom)
points for which
yrandom<y(xrandom). The value of the integral will
be the ratio of points within the integral times the total area sampled, or
(S/N) * h * (b-a).
Modeling Stochastic Phenomena
Monte Carlo modeling and probability.
Stochastic phenomena are phenomena heavily influenced by random events. An example might be
one's score in a game of darts. While skill is involved, randomness also plays a role. Many
physical phenomena fit into this category.
Tracking the change in any dynamic system generally is done by one of
two methods. The most common method (integration of differential equations)
is to determine general properties of the system, to specify the rate of change
over time, and to track the change in those general quantities over time. While
a powerful tool computationally, this requires the system being solved to be
easily represented by a few variables, and to not be strongly affected by random events.
Monte Carlo modeling attempts to solve dynamic problems by tracking individual elements,
in this case, individual predators and prey, and determining the occurence
of key events (such as eating and being eaten) by evaluation of the
probability of occurence of that event within a given time.
While more cumbersome computationally, this method not only allows us to
deal with the effects of random events on a small population more
realistically, but it also allows us to more easily track the change
of traits within the population.
Modeling using known probabilities
The simplest method of Monte Carlo modeling involves cases where there is
a known probability. Rolling dice, flipping coins, or spinning a spinner are examples.
Often, however, determining the probability of an event occuring can be the
difficult part of the problem, however, if you know the probability, and
can compute a random number, you can build the model.
Modeling using timescales
An important parameter in many Monte Carlo model is the timescale for events to occur.
The time scale is the "average" time between occurences. That means, that
in a large population, that in a period of the "average" time, about half
of the population will undergo the event in question. This is similar to
another scientific quantity, half-life, and we can state the probability
of the event occuring for an individual within the population during a
period of time t as
P(t) = (1-(1/2)t/t1/2)
The mode progresses by stepping forward some time t, determining if an event
has happened by evaluating the probability of the event occuring (on a scale
of 0 to 1), producing a pseudo-random number between 0 and 1, and checking to
see if the pseudo-random number is less than the probability of the event occuring.