Strong approximations in the almost sure central limit theorem and limit behavior of the center of mass

Zhishui Hu, Wei Wang, Liang DongStochastic 2025 · DOI

TL;DR

This paper extends the "almost sure central limit theorem" (ASCLT) to more general sequences of random variables, moving beyond independent and identically distributed (i.i.d.) cases. It also investigates the law of the iterated logarithm for the "center of mass" of these sequences, with applications to step-reinforced random walks.

What Problem Does This Paper Address?

The "almost sure central limit theorem" (ASCLT) is a powerful result that describes the asymptotic behavior of sums of random variables. While it's well-established for i.i.d. sequences, extending it to more complex dependent structures, like martingales or sequences with specific approximation conditions, is crucial for a broader understanding of random processes. Furthermore, understanding the long-term behavior of the "center of mass" of such sequences is important in various fields, from physics (e.g., polymer chains) to finance. This paper aims to bridge these gaps by providing strong approximation results for ASCLTs and the law of the iterated logarithm for the center of mass under more general conditions.

Methods

The authors employ "strong approximation" techniques, which involve approximating a sequence of random variables with a related sequence that is "close" to a known process, often a Brownian motion. This allows them to leverage the well-understood properties of Brownian motion to derive results for the original sequence. Key tools include:

  • Almost Sure Central Limit Theorem (ASCLT): The paper establishes new ASCLT results for sequences satisfying specific "strong approximation conditions." This means that for a sequence of random variables $M_n$, if $M_n$ is "close" to a Brownian motion $W(b_n)$ in a specific way, then a scaled version of $M_n$ converges in distribution to a normal distribution almost surely.
  • Law of the Iterated Logarithm (LIL): The paper extends LIL results to the "center of mass" of random sequences. The center of mass $G_n$ is essentially the average of the sequence up to time $n$. The LIL provides bounds on how fast the fluctuations of $G_n$ grow.
  • Step-Reinforced Random Walks: The theoretical results are applied to analyze the asymptotic behavior of step-reinforced random walks, a class of random walks with memory.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Generalizing ASCLT: The paper shows that ASCLT can hold even when the random variables are not independent, as long as they satisfy certain "strong approximation" conditions relative to a Brownian motion. This broadens the applicability of ASCLT.
  • LIL for Center of Mass: The authors establish LIL for the center of mass of general random sequences, which is a significant extension of existing results for simpler random walks.
  • Applications to Step-Reinforced Walks: The theoretical framework is successfully applied to step-reinforced random walks, providing concrete ASCLT and LIL results for these models, which have connections to real-world phenomena.
  • Condition (1.8) is Necessary: The paper suggests that the condition $M_n - W(b_n) = o(\sqrt{b_n})$ a.s. (Eq. 1.8) is crucial for the ASCLT to hold in its general form, highlighting its importance.

Key Results

  • Theorem 1.1: Establishes an ASCLT for sequences $M_n$ that are strongly approximated by a Brownian motion $W(b_n)$ under specific conditions on $b_n$ and the approximation error.
  • Theorem 1.2: Derives the law of the iterated logarithm for the center of mass $G_n$ of a random sequence $M_n$ under a weaker strong approximation condition.
  • Theorem 2.1 & 2.3: Provide specific ASCLT and LIL results for positively and negatively step-reinforced random walks, respectively, based on the general theorems. For instance, for a positively step-reinforced random walk with parameter $p \in [0, 1/2)$, the paper gives the asymptotic behavior of the center of mass $G_n$.

Limitations and Future Work

  • Moment Conditions: The proofs often rely on moment conditions (e.g., $E|X_1|^{2+\delta} < \infty$). Weakening these conditions could lead to more general results.
  • Specific Approximation Conditions: The theorems are contingent on specific "strong approximation conditions." Exploring the exact boundary of these conditions and their necessity is an ongoing area of research.
  • Further Applications: The authors suggest that their theorems can be applied to other types of dependent random sequences, which would be a natural direction for future work.

Summary updated

BibTeX
@inproceedings{hu2025strong,
  author = {Zhishui Hu and Wei Wang and Liang Dong},
  title = {{Strong approximations in the almost sure central limit theorem and limit behavior of the center of mass}},
  booktitle = {Stochastic},
  year = {2025},
  doi = {10.1016/j.spa.2025.104570},
}
Figure