Counting Process Simulation
Simulation idea
We want to simulate a counting process over a time interval T, where events happen:
- Independently
- Uniformly in time
- With constant average rate λ
We approximate this by dividing T into n small intervals. For each interval:
$$\text{Probability of an event} = \frac{\lambda}{n}$$
This is equivalent to simulating a Bernoulli process in discrete time.
Theoretical Properties
- Limiting Process:
As \(n \to \infty\), this discrete process converges to a Poisson process N(t)with rate λ. - Properties of Poisson Process:
- N(t) counts the number of events up to time t
- Number of events in disjoint intervals are independent
- Number of events in interval of length t follows Poisson distribution: $$P(N(t) = k) = \frac{(\lambda t)^k e^{-\lambda t}}{k!}$$
- Inter-event times (time between events) are exponentially distributed with mean \(1/\lambda\)
- Interpretation of λ:
- Average number of events per unit time
- Higher λ → more frequent events
Events occur independently and uniformly over time with average rate λ.