🎯Stationary Distribution

NEW

Also known as: Invariant Distribution

A probability distribution that remains unchanged after applying a Markov transition kernel or matrix.

DESCRIPTION

A stationary distribution π satisfies π = πP for a Markov chain with transition matrix P. It describes the long-run fraction of time spent in each state when the chain is ergodic. Stationary distributions are central to MCMC, PageRank, and stochastic modeling. Under irreducibility and aperiodicity, finite-state chains have a unique stationary distribution. Reversibility and detailed balance offer constructive ways to design chains with desired stationary targets. In practice, stationary distributions enable steady-state analysis and sampling.

ALTERNATIVE DEFINITIONS

Wikipedia

In the study of Markov chains, a stationary distribution is a probability distribution that remains unchanged under the operation of the transition probabilities of the chain.

Frames stationary distribution specifically in the Markov chain context.

Source

Wolfram MathWorld

A stationary distribution of a Markov chain is a probability distribution that is unchanged by the transition matrix.

Concise linear algebra characterization via the transition matrix.

Source

StatLect

A stationary distribution is a distribution that does not change when the Markov chain evolves, i.e., it is an invariant distribution.

Explicitly equates stationary with invariant distribution.

Source

RELATED TERMS

Markov Chain Invariant Measure Detailed Balance Reversibility Mixing Time Ergodicity Transition Matrix PageRank Metropolis-Hastings Gibbs Sampling

COMPETING TERMS

Transient Distribution

KNOWLEDGE AREAS

Mathematics Data Science AI

TARGET AUDIENCE

Data Scientist AI Engineer

EXTERNAL SOURCES

• Wikipedia - Stationary distribution
• MIT OpenCourseWare - Lecture on Markov matrices and steady states
• StatLect - Detailed explanation with examples
• Wolfram MathWorld - Definition and properties of stationary distributions

TOP QUESTIONS

What is a Stationary Distribution in Markov chains?

It’s a distribution π satisfying π = πP. If the chain is ergodic, the empirical distribution over time converges to π.

Category: Definition

How is a Stationary Distribution used in MCMC?

MCMC algorithms are designed so the target distribution is stationary for the chain. Sampling the chain yields draws from the target under ergodicity.

Category: Application

When is a Stationary Distribution unique?

For finite-state chains, irreducibility and aperiodicity imply a unique stationary distribution. Without these, multiple or none may exist.

Category: Requirements

What is detailed balance and why does it matter?

Detailed balance ensures reversibility, making π an invariant distribution. It simplifies constructing chains with a desired stationary distribution.

Category: Definition

How do you compute a Stationary Distribution?

Solve π = πP with normalization constraints. For large systems, power iteration or simulation estimates are common.

Category: Example

What mistakes occur when identifying Stationary Distributions?

Confusing transient behavior with stationary states or ignoring reducibility and periodicity leads to wrong conclusions.

Category: Mistakes

What’s the risk of assuming Stationarity without checking?

Non-stationary dynamics yield biased long-run estimates and misinterpretations. Diagnostics and theoretical conditions must be verified.

Category: Risks

How is Stationary Distribution related to PageRank?

PageRank computes the stationary distribution of a random surfer model. It represents long-run visitation probabilities to pages.

Category: Example

What is a competing concept to Stationary Distribution?

Transient distributions describe short-term state probabilities that change over steps, contrasting stationary behavior.

Category: Alternatives