Probability Density Function

Probability Density Function
💡No image available
Overview
Defines	Probabilities via integration over intervals
Used in	Continuous probability distributions
Synonyms	Density function, distribution density

A probability density function (PDF) is a function used in probability theory and statistics to describe the likelihood distribution of a continuous random variable. The value of the PDF at a given point indicates the relative density of probability rather than a probability itself; probabilities for continuous variables are obtained by integrating the PDF over an interval. PDFs are fundamental to topics such as continuous random variables, integration, and statistical inference.

Definition

Let (X) be a real-valued random variable. If (X) is continuous and has a distribution that is absolutely continuous with respect to Lebesgue measure, then there exists a nonnegative function (f) such that for any interval ([a,b]), [ \Pr(a \le X \le b)=\int_a^b f(x),dx. ] In this setting, (f) is called the probability density function of (X).

A PDF must satisfy two core properties: (1) (f(x)\ge 0) for all (x), and (2) the total probability is 1, meaning (\int_{-\infty}^{\infty} f(x),dx=1). Because (X) is continuous, (\Pr(X=x)=0) for any exact value (x), even when (f(x)) is positive; probabilities are associated with ranges, not single points.

Relationship to the cumulative distribution function

The cumulative distribution function (CDF) of (X), often written (F(x)=\Pr(X\le x)), is related to the PDF by [ F(x)=\int_{-\infty}^{x} f(t),dt ] for continuous distributions with a well-defined density. Conversely, when (F) is differentiable at (x), the PDF can be recovered as (f(x)=F'(x)).

This relationship clarifies the difference between density and probability. The CDF maps a point to the probability of being less than or equal to that point, while the PDF is a rate or density that must be integrated over an interval to obtain the corresponding probability.

Common examples

Many widely used distributions in statistics and science are defined by explicit PDFs. For example, the normal distribution has a PDF given by the familiar bell curve, and it is often used to model measurement errors and other approximately Gaussian phenomena. The exponential distribution provides a model for waiting times in a Poisson process, using a PDF that decays exponentially.

Other examples include the uniform distribution on an interval, where the PDF is constant over the support, and distributions such as the gamma distribution, whose PDF generalizes the exponential case. In applications, these PDFs are used to compute probabilities, expected values, and tail probabilities by applying integral calculations.

Properties and computation

Given a PDF (f) for a continuous random variable (X), the probability of an event defined through inequalities is computed using integrals over the relevant region. For instance, (\Pr(X\ge c)=\int_c^\infty f(x),dx) for any real (c). The expected value of a function (g(X)) can similarly be computed as [ \mathbb{E}[g(X)] = \int_{-\infty}^{\infty} g(x) f(x),dx, ] whenever the integral exists. This connects PDFs to moment generating function methods and other tools used in probability and statistics.

In numerical work, PDFs are often approximated using quadrature methods or sampled through Monte Carlo simulation. In inferential settings, PDFs also underpin likelihood-based approaches, where the likelihood function is constructed from PDFs to fit model parameters.

Measure-theoretic viewpoint

In a measure-theoretic formulation, a PDF is best understood through the concept of absolute continuity: a distribution is represented by a density function with respect to a base measure (typically Lebesgue measure on (\mathbb{R})). This viewpoint helps clarify when a density exists and how to distinguish purely continuous distributions from mixed or discrete ones.

The connection to Radon–Nikodym derivative formalizes the idea that, when it exists, the PDF is the Radon–Nikodym derivative of the distribution measure of (X) with respect to Lebesgue measure. This framework unifies the treatment of probabilities across discrete, continuous, and mixed models and is widely used in advanced probability theory.