You are on page 1of 2

Alex Chen

Workshop on models of an epidemic


1. Here is the graph of the population over time when split into the susceptible, infected and
recovered populations.

At the beginning, since I is small, the initial growth rate of the Infected should be doubling every week, since
a = 2ln(2). Once the Susceptible fall to half of the population, more people are recovered or infected than
not, so the number of people that can be infected becomes small, and so number of infected starts to
𝑑𝐼 𝑎
decrease. You can also see this by considering when 𝑑𝑡 = 0, which happens when 𝑁 𝐼𝑆 = 𝑏𝐼 which can be
𝑁
rearranged to give 𝑆 = 2 , so again we see the turning point is when the number of Susceptible falls to half of
the overall population.
The code for this graph is attached in the file ContinuousSIR.m
2. The code for the stochastic model is attached as stochasticSIR.m
3. I don’t have a figure for my code, even though my code appears exactly the same as my partner’s, but
perhaps running the code on another computer will show that the results are very similar to the deterministic
model.
6. With the continuous approach, it gives a good general shape, but the number of people who are
susceptible, infected or recovered at any time becomes a continuous value, rather than integer values as they
would be in real life. The continuous model is also helpful as we can analyze the functions in order to
extrapolate. However, it is deterministic and describes a population using statistics taken as an average from
the whole population.
Alex Chen

With the stochastic approach, the fact that at any point in time, the values are all integers and that there is an
element of chance modeled by the random number means that it is more representative of the non-
deterministic nature of epidemics. However, the probabilities are dependent on dt, the unit time, so this must
be chosen carefully, and if the values of the probabilities are taken from the continuous approach, then this
model will simply provide an estimate of the continuous approach over a large number of runs.
With the individual agent model, it is difficult to generalize from a small number of runs of the program, and
it is difficult to perform analysis to have some idea of the features of the system. However, this model
provides the closest analogy of what actually happens in real life, in the sense that infections are spread by
individuals who come into contact, not just over time randomly as suggested by the previous two models. We
can also model far more complex and dynamic situations with the individual agent model.

You might also like