Exponential

Definition

Concept

Exponential distribution models waiting times between events in a homogeneous Poisson process. Continuous, non-negative support. Describes time until next event.

Support

Domain: x ∈ [0, ∞). Values represent elapsed time or distance between occurrences.

Historical Context

Introduced in early 20th century to describe radioactive decay intervals, reliability lifetimes, and queuing delays.

Properties

Memorylessness

Unique continuous distribution with P(X > s + t | X > s) = P(X > t). No aging effect.

Monotonicity

Decreasing failure rate function: hazard rate constant over time.

Skewness and Kurtosis

Right-skewed distribution. Skewness = 2, excess kurtosis = 6.

Probability Density Function and Cumulative Distribution Function

Probability Density Function (PDF)

Defines likelihood of specific event time.

f(x; λ) = λ e-λx, for x ≥ 0, λ > 0

Cumulative Distribution Function (CDF)

Probability event occurs by time x.

F(x; λ) = 1 - e-λx, for x ≥ 0

Survival Function

Probability event time exceeds x.

S(x) = P(X > x) = e-λx

Parameters

Rate Parameter (λ)

λ > 0 controls event frequency. Higher λ → shorter waiting times.

Scale Parameter (θ)

θ = 1/λ, scale of time between events.

Parameter Roles

λ determines shape and scale implicitly. Single-parameter distribution.

Mean and Variance

Expected Value (Mean)

Mean waiting time: E[X] = 1/λ.

Variance

Var(X) = 1/λ², showing dispersion scales with λ.

Higher Moments

n-th moment: E[Xⁿ] = n!/λⁿ, factorial moments increase rapidly.

Memoryless Property

Definition

Conditional probability independent of elapsed time: P(X > s + t | X > s) = P(X > t).

Implications

System "age" irrelevant. No accumulated wear or fatigue.

Uniqueness

Only exponential among continuous distributions has this property.

P(X > s + t | X > s) = P(X > t)→ S(s + t) / S(s) = S(t)→ e-λ(s + t) / e-λs = e-λt

Applications

Reliability Theory

Models component lifetimes, failure rates constant over time.

Queuing Theory

Interarrival and service times in M/M/1 queues.

Physics and Biology

Radioactive decay, molecular reaction times, neuron firing intervals.

Telecommunications

Modeling packet arrivals, call durations.

Finance

Modeling time between trades or events in high-frequency trading.

Relation to Other Distributions

Poisson Process

Interarrival times between Poisson events are exponential.

Gamma Distribution

Exponential is special case of gamma with shape = 1.

Weibull Distribution

Weibull generalizes exponential with shape parameter ≠ 1.

Geometric Distribution

Discrete analogue of exponential distribution.

Relation to Erlang Distribution

Erlang sums independent exponential variables with integer shape.

Parameter Estimation

Maximum Likelihood Estimation (MLE)

λ̂ = 1 / sample mean.

Method of Moments

Match sample mean to theoretical mean 1/λ.

Bayesian Estimation

Conjugate prior: gamma distribution on λ.

Confidence Intervals

Based on chi-square distribution for sum of exponentials.

Bias and Consistency

MLE unbiased and consistent for large samples.

Given sample {x₁, x₂, ..., xₙ}:λ̂ = n / ΣxᵢLog-likelihood: L(λ) = n log λ - λ Σxᵢ

Simulation Techniques

Inverse Transform Sampling

Generate uniform U ~ U(0,1), then X = -ln(U)/λ.

Rejection Sampling

Less efficient for exponential; rarely used.

Use in Monte Carlo Methods

Generate event times in stochastic simulations.

Software Implementations

Available in R (rexp), Python (numpy.random.exponential), MATLAB.

Random Number Generator Quality

Quality of uniform RNG affects exponential sample accuracy.

Algorithm:1. Generate U ~ Uniform(0,1)2. Compute X = - (1/λ) * ln(U)3. Return X as exponential random variable

Examples

Radioactive Decay

Time between decays of unstable atoms follows exponential distribution.

Customer Service

Waiting times between arrivals at a service desk.

System Failures

Time to failure for electronic components with constant hazard rate.

Call Center

Intervals between incoming calls modeled as exponential.

Network Packets

Time gaps between packet arrivals on network routers.

Limitations

Constant Hazard Rate Assumption

Does not model aging or wear-out effects.

Lack of Flexibility

Single parameter limits shape adaptation.

Not Suitable for Multi-modal Data

Cannot represent distributions with multiple peaks.

Over-Simplification

Real-world event times may exhibit dependencies violating memorylessness.

Alternatives

Weibull, log-normal, and gamma distributions for more complex modeling.

References

• Ross, S. M. "Introduction to Probability Models," Academic Press, 11th Edition, 2014, pp. 58-65.
• Feller, W. "An Introduction to Probability Theory and Its Applications," Wiley, Vol. 1, 1968, pp. 238-242.
• Lawless, J. F. "Statistical Models and Methods for Lifetime Data," Wiley, 2003, pp. 20-35.
• Papoulis, A., and Pillai, S. U. "Probability, Random Variables, and Stochastic Processes," McGraw-Hill, 4th Edition, 2002, pp. 260-270.
• Gross, D., and Harris, C. M. "Fundamentals of Queueing Theory," Wiley, 4th Edition, 2008, pp. 45-50.