[Paper Review] On spectral properties and statistical analysis of Fisher-Snedecor diffusion

Florin Avram, Nikolai Leonenko+1 more|arXiv (Cornell University)|Jul 28, 2010

Stochastic processes and statistical mechanics67 references3 citations

TL;DR

This paper introduces the Fisher-Snedecor diffusion, an ergodic diffusion process with a Fisher-Snedecor invariant distribution, and derives its spectral representation via a finite set of orthogonal eigenfunctions (Fisher-Snedecor polynomials) and a continuous spectral component. It proposes consistent, asymptotically normal moment-based estimators and a Stein equation-based statistical test for marginal distributional assumptions.

ABSTRACT

We consider the problem of parameter estimation for an ergodic diffusion with Fisher-Snedecor invariant distribution, to be called Fisher-Snedecor diffusion. We compute the spectral representation of its transition density, which involves a finite number of discrete eigenfunctions (Fisher-Snedecor polynomials) as well as a continuous part. We propose moments based estimators (related to the Fisher-Snedecor polynomials) and prove their consistency and asymptotic normality. Furthermore, we propose a statistical test for the distributional assumptions on the marginal distribution of the Fisher-Snedecor diffusion, based on the moment condition derived from the corresponding Stein's equation.

Motivation & Objective

• To develop a statistical framework for parameter estimation in ergodic diffusion processes with Fisher-Snedecor invariant distributions.
• To derive the spectral representation of the transition density of the Fisher-Snedecor diffusion, including discrete eigenfunctions and a continuous spectral measure.
• To construct moment-based estimators grounded in the eigenfunction expansion and prove their consistency and asymptotic normality.
• To design a statistical test for the marginal distributional assumption using moment conditions derived from the Stein equation associated with the invariant distribution.

Proposed method

• Derives the spectral decomposition of the transition density using Sturm-Liouville theory, identifying eigenfunctions as Fisher-Snedecor polynomials.
• Applies the Liouville transformation to reduce the Fokker-Planck equation to a canonical form, enabling explicit spectral analysis.
• Constructs moment-based estimators by matching empirical moments of the process to theoretical moments derived from the eigenfunction expansion.
• Utilizes the Stein equation associated with the Fisher-Snedecor distribution to derive moment conditions for goodness-of-fit testing.
• Establishes the asymptotic normality of estimators through martingale limit theory and spectral properties of the generator.
• Analyzes boundary behavior of the associated Sturm-Liouville problem to confirm the nature of the spectrum (discrete and continuous components).

Experimental results

Research questions

• RQ1 What is the spectral representation of the transition density for a diffusion process with Fisher-Snedecor invariant distribution?
• RQ2 How can consistent and asymptotically normal moment-based estimators be constructed for the parameters of the Fisher-Snedecor diffusion?
• RQ3 What moment conditions derived from the Stein equation can be used to test the validity of the Fisher-Snedecor distributional assumption?
• RQ4 How do the spectral properties of the generator influence the statistical behavior of the diffusion process?
• RQ5 What is the role of the Fisher-Snedecor polynomials in the eigenfunction expansion and estimation framework?

Key findings

• The transition density of the Fisher-Snedecor diffusion admits a spectral representation involving a finite number of discrete eigenfunctions (Fisher-Snedecor polynomials) and a continuous spectral component.
• The moment-based estimators for the parameters of the Fisher-Snedecor diffusion are consistent and asymptotically normal under regularity conditions.
• The Fisher-Snedecor polynomials arise as eigenfunctions of the generator of the diffusion process and are orthogonal with respect to the invariant measure.
• A statistical test for the marginal distributional assumption is constructed using moment conditions derived from the Stein equation of the Fisher-Snedecor distribution.
• The boundary classification of the associated Sturm-Liouville problem confirms the spectral properties: the left boundary is of LC type for α < 4 and LP type for α > 4, while the right boundary is always LP type.
• The Wronskian of the fundamental solutions in the Liouville normal form is constant, confirming the spectral structure and enabling explicit computation of the spectral measure.