[Paper Review] Some Characterizations Of Mixed Renewal Processes

Demetrios P. Lyberopoulos, N. D. Macheras|arXiv (Cornell University)|May 20, 2012

advanced mathematical theories9 references6 citations

TL;DR

This paper provides new characterizations of mixed renewal processes (MRPs) using exchangeability and disintegration theory, establishing a rigorous foundation for their construction. It proves that MRPs can be equivalently defined via conditional independence and disintegrating measures, and demonstrates that Huang's and the authors' definitions of MRPs coincide under typical application conditions, with explicit constructions and measure computations provided for concrete examples.

ABSTRACT

Some characterizations of mixed renewal processes in terms of exchangeability and of different types of disintegrations are given. As a consequence, an existence result for mixed renewal processes, providing also a new construction for them, is obtained. As an application, some concrete examples of constructing such processes are presented and the corresponding disintegrating measures are explicitly computed.

Motivation & Objective

• To establish a new, structurally sound definition of mixed renewal processes (MRPs) that explicitly incorporates the parameter Θ, essential for risk-theoretic applications.
• To investigate the role of disintegration in MRPs by introducing and analyzing product and subfield regular conditional probabilities.
• To clarify the relationship between two existing definitions of MRPs—one by the authors and one by Huang—showing their equivalence under typical application conditions.
• To provide a constructive existence result for MRPs by explicitly computing disintegrating measures in concrete examples.
• To demonstrate the necessity of key assumptions in theoretical results through counterexamples, ensuring robustness of the characterizations.

Proposed method

• Introduces a new definition of MRPs (Definition 3.1) based on conditioning on a structural parameter Θ, aligning with mixed Poisson processes.
• Applies disintegration theory to decompose the probability measure into conditional distributions, using product and subfield regular conditional probabilities.
• Employs the concept of conditionally i.i.d. families of random variables (Proposition 3.11) to extend prior results on conditional independence.
• Uses exchangeability criteria via disintegration (Theorem 4.4) to derive characterizations of MRPs under different assumptions.
• Constructs non-trivial probability spaces supporting MRPs (Example 5.5), extending prior constructions for mixed Poisson processes.
• Computes disintegrating measures explicitly for specific examples, including cases with Gamma and inverse Gaussian components.

Experimental results

Research questions

• RQ1 How can mixed renewal processes be characterized using exchangeability and disintegration theory?
• RQ2 What conditions ensure the equivalence between the authors' definition of MRP and Huang's definition?
• RQ3 What is the role of the structural parameter Θ in the construction and characterization of MRPs?
• RQ4 Are the assumptions in the main theorems (e.g., conditional i.i.d. or identically distributed interarrival times) necessary, or can they be relaxed?
• RQ5 Can explicit constructions of MRPs be achieved with computable disintegrating measures for practical applications?

Key findings

• The authors' definition of MRP (Definition 3.1) is shown to be natural and consistent, as it reduces to ordinary renewal processes under disintegrating measures.
• Theorem 4.9 establishes that under typical application conditions, the two definitions of MRPs (by the authors and Huang) are equivalent, validating the broader applicability of the framework.
• Explicit disintegrating measures are computed for a class of MRPs involving Gamma-distributed Θ and inverse Gaussian interarrival times, with $ P_{ heta}( ilde{B}) = igotimes_{n} Q_n( heta)( ilde{C}_k) $ and $ P( ilde{B} imes E) = rac{ u^{ u}}{ u! u} imes ext{integral expression} $.
• Example 5.7 shows that the assumption of $ P_{ heta} $-i.i.d. interarrival times is essential: without it, even conditional independence fails to imply exchangeability, as demonstrated by $ P(W_1 eq 2, W_2 eq 1) \neq P(W_1 eq 1, W_2 eq 2) $.
• The paper proves that $ \{W_n\} $ is not $ P $-exchangeable in Example 5.7, despite being $ P_{\theta} $-independent for each $ \theta > 0 $, highlighting the subtlety of exchangeability in mixed processes.
• The construction in Example 5.5 provides a new method for building non-trivial probability spaces supporting MRPs, extending prior work on mixed Poisson processes.