Exponential Distribution

Core Idea

The exponential distribution with rate parameter λ > 0 has PDF f(x) = λe^-λx for x ≥ 0, and models waiting times until an event when events occur at a constant rate λ. Mean is 1/λ and variance is 1/λ². The exponential distribution is memoryless: P(X > s + t | X > s) = P(X > t), meaning remaining time doesn't depend on elapsed time. It naturally arises as the continuous analog of the geometric distribution.

How It's Best Learned

Derive memoryless property algebraically. Model real waiting time scenarios (customer service, radioactive decay). Relate to Poisson processes.

Common Misconceptions

Confusing rate λ with scale parameter 1/λ. Not recognizing memorylessness property. Applying exponential without constant rate assumption.

Explainer

From continuous random variables, you know that a continuous distribution is described by a probability density function (PDF) f(x), where areas under the curve give probabilities. The exponential distribution is one of the simplest and most widely applicable: its PDF is f(x) = λe^−λx for x ≥ 0, and zero for x < 0. The parameter λ (lambda) is the rate — how many events occur per unit time on average. If calls arrive at a call center at a rate of 5 per hour, then λ = 5 and the waiting time between successive calls follows Exp(5). The mean waiting time is 1/λ = 1/5 of an hour = 12 minutes. Notice the inverse relationship: a higher rate means shorter waits on average.

The memoryless property is what makes the exponential distribution unique among continuous distributions. It states that P(X > s + t | X > s) = P(X > t). In English: if you have already waited s minutes with no call, the probability you will wait at least t more minutes is exactly the same as if you had just started waiting. Past waiting time gives you no information about future waiting time. The mathematical proof is direct from the CDF: P(X > x) = e^−λx, so P(X > s + t | X > s) = P(X > s + t) / P(X > s) = e^(−λ(s+t)) / e^−λs = e^−λt = P(X > t). This is analogous to the geometric distribution's memoryless property for discrete waiting times — the exponential is its continuous cousin.

Memorylessness is both the strength and the limitation of the exponential model. It is the right model when the event has no "aging" or "wear" — a radioactive atom is equally likely to decay in the next second regardless of how long it has already existed; a packet in a network router is equally likely to depart in the next millisecond regardless of how long it has been queued. But it is the wrong model for things that do age: a machine that is more likely to fail the older it gets is better modeled by a Weibull distribution, which generalizes the exponential by allowing the hazard rate to increase over time.

The exponential distribution is deeply connected to the Poisson process. If events occur at a constant rate λ (a Poisson process), then the number of events in a fixed time interval follows a Poisson distribution, and the waiting time between consecutive events follows Exp(λ). These two distributions are two sides of the same underlying process: Poisson counts events in time, exponential measures gaps between them. When you see a Poisson random variable in a problem, the inter-arrival times are automatically exponential — and when you see exponential waiting times, you can count arrivals with a Poisson distribution. This pairing makes the exponential distribution central to queueing theory, reliability engineering, and any stochastic model where events occur unpredictably at a steady background rate.