The beta distribution is a family of continuous probability distributions defined on the interval $[0, 1]$ or $(0, 1)$, parameterized by two positive shape parameters, commonly denoted by $\alpha$ and $\beta$. These parameters can be interpreted as the number of successes and failures respectively in a sequence of Bernoulli trials, making the beta distribution particularly useful for modeling proportions and probabilities.
The beta distribution serves as the conjugate prior for the Bernoulli, Binomial, Negative Binomial, and Geometric distributions in Bayesian inference. This property allows for analytically tractable posterior distributions, which is why it is a cornerstone in Bayesian machine learning and A/B testing frameworks.
Definition
A random variable $X$ follows a beta distribution if its probability density function (PDF) is given by:
where $x \in [0, 1]$, $\alpha > 0$, $\beta > 0$, and $B(\alpha, \beta)$ is the Beta function, which serves as a normalization constant to ensure the PDF integrates to 1.
Here, $\Gamma(\cdot)$ denotes the Gamma function, which generalizes the factorial function to real and complex numbers.
Cumulative Distribution Function
The cumulative distribution function (CDF) of the beta distribution is the regularized incomplete beta function $I_x(\alpha, \beta)$:
Parameters & Shapes
The shape of the beta distribution is heavily influenced by the parameters $\alpha$ and $\beta$. Both parameters must be strictly positive. The distribution can take on a wide variety of shapes, including uniform, U-shaped, unimodal, and skewed forms.
| Condition | Mode | Shape Description |
|---|---|---|
| $\alpha = \beta = 1$ | Undefined (Uniform) | Uniform distribution on $[0, 1]$. Every value is equally likely. |
| $\alpha = \beta > 1$ | $0.5$ | Symmetric, bell-shaped around 0.5. Approaches normal as $\alpha$ increases. |
| $\alpha = \beta < 1$ | 0 and 1 | U-shaped (bathtub). High probability at extremes, low in the center. |
| $\alpha > \beta$ | $> 0.5$ | Skewed right (negative skew). Mass concentrated toward 1. |
| $\alpha < \beta$ | $< 0.5$ | Skewed left (positive skew). Mass concentrated toward 0. |
Moments
The moments of the beta distribution are well-defined and can be expressed in terms of $\alpha$ and $\beta$. For a random variable $X \sim \text{Beta}(\alpha, \beta)$:
Relationships with Other Distributions
- Binomial: The beta distribution is the conjugate prior for the binomial distribution. If $X \sim \text{Beta}(\alpha, \beta)$ is the prior for a success probability $p$, and we observe $k$ successes in $n$ trials, the posterior is $\text{Beta}(\alpha + k, \beta + n - k)$.
- Dirichlet: The beta distribution is a special case of the Dirichlet distribution when there are only two categories ($K=2$).
- Normal: As $\alpha, \beta \to \infty$ with their ratio fixed, the beta distribution converges to a normal distribution (appropriately scaled).
- Arcsine: When $\alpha = \beta = 0.5$, the beta distribution becomes the arcsine distribution.
Applications
The beta distribution is extensively used across various fields due to its flexibility and bounded support:
- Bayesian Statistics: Used as a prior distribution for probabilities and proportions. Its conjugacy properties simplify posterior updates.
- Project Management: In the Program Evaluation and Review Technique (PERT), the beta distribution is used to model the uncertainty of activity durations, typically parameterized by optimistic, pessimistic, and most likely estimates.
- Machine Learning: Used in reinforcement learning (e.g., Thompson sampling) to balance exploration and exploitation in multi-armed bandit problems.
- Biology & Genetics: Modeling allele frequencies in population genetics and gene expression data.
- Finance: Modeling default probabilities and credit risk assessment.
References
- Congdon, P. (2007). Bayesian Statistical Methods. Wiley.
- Kotz, S., & Johnson, N. L. (1981). Continuous Univariate Distributions, Vol 1. Wiley.
- Wikipedia. Beta distribution. Retrieved 2025.