Modes of Convergence

Sequences of random variables can converge in several inequivalent senses, almost surely, in probability, in the mean of order p, and in distribution, and understanding their hierarchy is essential to stating and proving every limit theorem precisely.

Definition

Modes of convergence are the distinct senses in which a sequence of random variables or their distributions can approach a limit, ranging from the strong almost-sure and mean convergence of the variables themselves to the weak convergence of their distributions.

Scope

The topic covers almost-sure convergence, convergence in probability, convergence in the p-th mean, and convergence in distribution, the implications and counterexamples relating them, uniform integrability as the bridge between convergence in probability and in mean, the portmanteau characterization of weak convergence, and tightness with Prohorov's theorem for relative compactness of families of measures.

Core questions

• What are the main senses in which random variables converge, and how do they differ?
• Which modes of convergence imply which others, and where do the implications fail?
• What additional condition upgrades convergence in probability to convergence in mean?
• When does a family of distributions have a convergent subsequence?

Key concepts

• almost-sure convergence
• convergence in probability
• convergence in mean
• weak convergence
• tightness and Prohorov's theorem

Key theories

Hierarchy of convergence modes

Almost-sure convergence and convergence in the p-th mean each imply convergence in probability, which in turn implies convergence in distribution, while the reverse implications generally fail, so the modes form a strict hierarchy with standard counterexamples.

Portmanteau theorem

Weak convergence of probability measures is equivalent to several conditions at once, including convergence of expectations of bounded continuous functions and convergence of the distribution function at every continuity point, giving flexible criteria for proving convergence in distribution.

Prohorov's theorem and tightness

A family of probability measures is relatively compact for weak convergence if and only if it is tight, meaning mass does not escape to infinity, which is the standard tool for extracting convergent subsequences in the study of limit theorems and stochastic processes.

Clinical relevance

Precise modes of convergence underlie the rigorous statements of consistency and asymptotic distribution in statistics, the convergence of simulation and approximation schemes, and the functional limit theorems, such as Donsker's invariance principle, that justify approximating complex stochastic systems by Brownian motion.

History

The careful distinction among modes of convergence emerged with the measure-theoretic foundations of probability, and the theory of weak convergence of measures on metric spaces, with tightness and Prohorov's compactness criterion, was systematized by Prohorov and Billingsley in the mid-twentieth century to support limit theorems for stochastic processes.

Key figures

• Patrick Billingsley
• Yuri Prohorov
• Aleksandr Khinchin

Seminal works

• billingsley1999convergence

Frequently asked questions

Why distinguish so many kinds of convergence?

Different limit theorems naturally yield different modes; the law of large numbers gives almost-sure convergence, the central limit theorem gives convergence in distribution, and conclusions about averages of the variables require convergence in mean, so the precise mode matters for what can be concluded.

What is tightness?

A family of distributions is tight if, for any required level, a single compact set carries at least that much probability for every member of the family; tightness prevents probability mass from leaking to infinity and is exactly the condition Prohorov's theorem needs for weak compactness.