Statistical Distributions Reference

Common probability distributions — PDFs, means, variances, and typical uses.

Reference Reference Updated Apr 19, 2026
Reference

Discrete

Name Support Mean Variance Used for
Bernoulli(p) {0, 1} p p(1−p) Single yes/no trial
Binomial(n, p) 0..n np np(1−p) Count of successes in n trials
Poisson(λ) 0, 1, … λ λ Rare-event counts (arrivals)
Geometric(p) 1, 2, … 1/p (1−p)/p² Trials until first success
Negative binomial 0, 1, … r(1−p)/p r(1−p)/p² Overdispersed counts
Uniform {1..n} 1..n (n+1)/2 (n²−1)/12 Discrete fair

Continuous

Name Support Mean Variance Used for
Uniform(a, b) [a, b] (a+b)/2 (b−a)²/12 Fair between bounds
Normal(μ, σ²) μ σ² CLT limit; measurement error
Log-normal (0, ∞) e^{μ+σ²/2} varies Positive-skewed quantities
Exponential(λ) [0, ∞) 1/λ 1/λ² Waiting time (memoryless)
Gamma(k, θ) [0, ∞) kθ² Sum of exponentials
Beta(α, β) [0, 1] α/(α+β) varies Proportion / rate
Chi-square(k) [0, ∞) k 2k Sum of squared normals
Student t(ν) 0 ν/(ν−2) Sample mean with unknown σ
F(ν₁, ν₂) [0, ∞) ANOVA, variance ratios
Cauchy undefined undefined Heavy-tailed
Weibull(k, λ) [0, ∞) varies varies Reliability / life

Rules of thumb

  • Normal approximation to binomial: np > 5 and n(1−p) > 5.
  • Poisson approximation to binomial: n large, p small, np moderate.
  • Exponential is memoryless — the only continuous distribution with this property.
  • Central Limit Theorem: sum of many IID finite-variance RVs → normal.

Last updated: