##  [Poisson Process](/poisson-process-0) 

 Definition

A point process on the real line (time) or in space in which counts in disjoint intervals are independent and the number of events in any interval of length t follows a Poisson distribution with mean λt for constant rate λ (homogeneous case).

 

 

 

 

 

 





## Principle

Principle

Independence of disjoint increments plus stationary increments (homogeneity) imply that event counts are Poisson-distributed with mean proportional to interval length; equivalently, interarrival times are independent exponential(λ) in the homogeneous temporal case.

 

 

 

 

 





## Demonstration

Demonstration

Situation: Calls arrive at a call center at an average rate λ per hour, assumed constant over observation. Recognition: Intervals without overlap are independent; the process is homogeneous. Action: Model arrivals as a Poisson process with rate λ and use exponential interarrival times to simulate or compute wait-time distributions. Consequence: The probability of k arrivals in t hours equals e^{-λt}(λt)^k/k!, and interarrival times are memoryless exponential with parameter λ.

 

 

 

 

## Misapplication

Misapplication

Applying the homogeneous Poisson process to data with time-varying rate (nonstationary) or dependent arrivals. The semantic error is assuming constant-rate stationarity and independence; doing so under time-varying intensity misestimates variances and event probabilities.

 

 

 

 

 





## Consequence

Consequence

Correct application yields closed-form probabilities for counts and waiting times, enabling tractable performance analysis and simulation. Incorrect application (ignoring nonstationarity or dependence) produces biased estimates of congestion, under- or overestimation of rare-event probabilities, and faulty capacity decisions.

 

 

 

 

## Reversal

Reversal

If the event rate λ(t) varies deterministically or stochastically over time (nonhomogeneous Poisson process) or events cluster (overdispersion), independence and exponential interarrival assumptions fail; one must use nonhomogeneous Poisson, Cox, renewal, or cluster process models.

 

 

 

 

 





## Boundary

Boundary

Clearly within: Photon counts detected from a constant-intensity source over short intervals with independent counts. Boundary case: Arrival data with slowly varying diurnal rate — over short windows a homogeneous Poisson may be acceptable, over long windows it is not. Clearly outside: Events exhibiting strong aftershock clustering (e.g., earthquakes with Omori-type clustering) which violate independence.

 

 

 

 

 





## Semantic Tension

Semantic Tension

Simplicity (tractable analytic form) ↔ Fidelity (ability to represent temporal heterogeneity or dependence): Poisson is analytically convenient but can misrepresent structured variability in real data.

 

 

 

 

 





## Synthesis

Synthesis

The Poisson process ties memoryless interarrival behavior to independent, stationary counts; recognizing when rate homogeneity or independence fails is critical because these assumptions determine closed-form probabilities and the appropriateness of exponential waiting-time reasoning.