On Goodness-of-Fit Testing for Volatility in McKean-Vlasov Models

Basic Information

• Paper ID: 2510.12607
• Title: On goodness-of-fit testing for volatility in McKean-Vlasov models
• Authors: Akram Heidari, Mark Podolskij (University of Luxembourg)
• Classification: stat.ME (Statistics - Methodology)
• Publication Date: October 14, 2025 (arXiv preprint)
• Paper Link: https://arxiv.org/abs/2510.12607

Abstract

This paper develops a statistical goodness-of-fit testing framework for volatility functions in McKean-Vlasov stochastic differential equations. McKean-Vlasov equations describe large-scale interacting particle systems with distribution-dependent dynamics. While integrated volatility estimation in classical SDEs is well-established, formal model validation and goodness-of-fit testing for McKean-Vlasov systems remain largely unexplored, particularly in the joint asymptotic regime of large particle limits and high-frequency sampling. The authors propose test statistics based on discrete observations from particle systems, analyzed under joint asymptotics where both particle numbers and sampling frequency increase simultaneously. The consistency of relevant estimators is established, and the test statistic is shown to satisfy a central limit theorem with convergence to a centered Gaussian law.

Research Background and Motivation

Problem Description

McKean-Vlasov stochastic differential equations are important mathematical tools for describing large-scale interacting particle systems, where the dynamics of each particle depend not only on its individual state but also on the statistical distribution of the entire system. This distribution dependence makes McKean-Vlasov models particularly suitable for capturing systemic interactions and emergent behavior.

Research Significance

1. Broad Applicability: McKean-Vlasov models have widespread applications in finance, physics, engineering, and other fields, including systemic risk modeling, mean-field games, and analysis of large-scale interacting systems
2. Theoretical Gap: While volatility estimation and testing theory for classical SDEs is mature, model validation methods for McKean-Vlasov systems remain lacking
3. Practical Need: In practical applications, rigorous verification of volatility function structure assumptions is necessary, as misspecification can significantly impact downstream predictions and risk measures

Limitations of Existing Methods

1. Inapplicability of Classical Methods: Existing SDE volatility testing methods (e.g., Dette & Podolskij, 2008) apply only to non-interacting systems
2. Insufficient Research: Existing McKean-Vlasov literature primarily focuses on parametric estimation of drift functions, with volatility testing essentially untouched
3. Methodological Gaps: Lack of statistical testing frameworks for handling distribution dependence and nonlinear effects

Core Contributions

1. Novel Framework: Proposes the first rigorous statistical framework for goodness-of-fit testing of volatility functions in McKean-Vlasov models
2. Dual Asymptotic Theory: Establishes theory under joint asymptotics with particle number N→∞ and sampling frequency increase (Δₙ→0)
3. Consistency Proofs: Proves consistency of involved estimators and central limit theorem for test statistics
4. Practical Testing Procedure: Constructs a testing procedure with correct asymptotic level and consistency against any fixed alternative hypothesis
5. Technical Innovation: Overcomes technical challenges posed by distribution dependence, nonlinearity, and path-dependent effects

Methodology Details

Problem Formulation

Consider an N-particle interacting system:

dX^i_t = b(X^i_t, μₜ)dt + a(X^i_t, μₜ)dW^i_t, i = 1,...,N, t ∈ [0,T]

where μₜ is the distribution of X^i_t, and the goal is to test whether the volatility function a(x,μ) belongs to a given parametric family.

Model Architecture

Hypothesis Testing Framework

Null Hypothesis:

H₀: L := min_{(λ₁,...,λₐ)∈ℝᵈ} ∫₀ᵀ ∫_ℝ (a²(x,μₜ) - Σᵈₖ₌₁ λₖa²ₖ(x,μₜ))² μₜ(dx)dt = 0

Alternative Hypothesis: H₁: L > 0

Test Statistic Construction

Closed-form expression for distance measure L:

L = B - (Γ₁,...,Γₐ)Λ⁻¹(Γ₁,...,Γₐ)ᵀ

where:

• B = ∫₀ᵀ ∫_ℝ a⁴(x,μₜ)μₜ(dx)dt
• Γₖ = ∫₀ᵀ ∫_ℝ a²ₖ(x,μₜ)a²(x,μₜ)μₜ(dx)dt
• Λₖ,ₗ = ∫₀ᵀ ∫_ℝ a²ₖ(x,μₜ)a²ₗ(x,μₜ)μₜ(dx)dt

Empirical Estimators

Based on discrete observations (X^i_{tⱼ}), construct estimators:

B̂ := 1/(3NΔₙ) Σᵢ₌₁ᴺ Σⱼ₌₁ⁿ |X^i_{tⱼ₊₁} - X^i_{tⱼ}|⁴

Γ̂ₖ := 1/N Σᵢ₌₁ᴺ Σⱼ₌₁ⁿ a²ₖ(X^i_{tⱼ}, μᴺ_{tⱼ})|X^i_{tⱼ₊₁} - X^i_{tⱼ}|²

Λ̂ₖ,ₗ := Δₙ/N Σᵢ₌₁ᴺ Σⱼ₌₁ⁿ a²ₖ(X^i_{tⱼ}, μᴺ_{tⱼ})a²ₗ(X^i_{tⱼ}, μᴺ_{tⱼ})

Final test statistic:

ŜN = B̂ - Γ̂ᵀΛ̂⁻¹Γ̂

Technical Innovations

1. Functional Derivatives: Employs functional derivatives to handle distribution dependence, a key technical tool for McKean-Vlasov equations
2. Dual Asymptotic Analysis: Simultaneously handles asymptotic behavior as N→∞ and Δₙ→0, requiring the balancing condition NΔ²ₙ→0
3. U-Statistic Decomposition: Uses Hoeffding decomposition techniques to address discrepancies between empirical and true distributions
4. Semimartingale Theory Application: Applies Itô formula and semimartingale properties for high-frequency statistical error estimation

Experimental Setup

Theoretical Verification Framework

This is primarily a theoretical work, with validity verified through mathematical proofs rather than traditional numerical experiments.

Key Assumptions

1. Assumption 1: Moment conditions on initial distributions
2. Assumption 2: Lipschitz continuity and linear growth conditions on coefficients
3. Assumption 3: Existence and smoothness of functional derivatives of volatility functions

Asymptotic Conditions

• Particle number N→∞
• Sampling interval Δₙ→0
• Balancing condition: NΔ²ₙ→0

Experimental Results

Main Theoretical Results

Theorem 4.1 (Consistency)

Under the stated assumptions:

√N(Λ̂ - Λ) = √NMΛ + oP(1)

Theorem 4.2 (Stochastic Expansion)

√N(Γ̂ₖ - Γₖ) = √NMₖ + oP(1)
√N(B̂ - B) = √NMB + oP(1)

Corollary 4.3 (Asymptotic Normality)

√N(ŜN - L) →^L N(0, τ²)

Testing Procedure

At significance level α, reject the null hypothesis when:

√NŜN/τ̂ > z₁₋α

where τ̂² is a consistent estimator of τ².

Theoretical Guarantees

1. Correct Asymptotic Level: The testing procedure achieves the correct α level under the null hypothesis
2. Consistency: For any fixed alternative H₁: L > 0, we have √NŜN →^P +∞
3. Relative Measure: Introduces standardized statistic G = L/B ∈ 0,1 for easier interpretation

Related Work

McKean-Vlasov Estimation Theory

• Amorino et al. (2024): Polynomial convergence rates for nonparametric estimation
• Belomestny et al. (2022): Semiparametric estimation
• Comte & Genon-Catalot (2024): Parametric inference
• Della Maestra & Hoffmann (2022): Nonparametric estimation

Classical SDE Testing Methods

• Dette & Podolskij (2008): Volatility testing for classical diffusion models
• Ait-Sahalia (1996): Continuous-time model testing
• Corradi & White (1999): Specification testing for diffusion variance

Advantages Relative to This Work

1. First Treatment of McKean-Vlasov: Existing methods apply only to classical SDEs
2. Distribution Dependence: Capable of handling volatility dependence on entire distributions
3. Dual Asymptotics: Jointly considers high-frequency and large-sample asymptotics

Conclusions and Discussion

Main Conclusions

1. Successfully establishes complete statistical theory for goodness-of-fit testing of volatility functions in McKean-Vlasov models
2. Proves consistency of estimators and asymptotic normality of test statistics under dual asymptotic framework
3. Constructs practical testing procedures with correct asymptotic properties

Limitations

1. Theoretical Work: Lacks numerical experiments to verify theoretical results
2. Assumption Conditions: Requires relatively strong smoothness and moment assumptions
3. Computational Complexity: Practical implementation requires computing functional derivatives, which may be complex
4. Finite Sample Properties: Does not provide finite-sample performance analysis

Future Directions

1. Numerical Verification: Validate theoretical results through Monte Carlo simulations
2. Practical Applications: Test method utility on financial data
3. Extensions: Generalize to multivariate settings and more general interaction structures
4. Computational Optimization: Develop efficient numerical algorithms

In-Depth Evaluation

Strengths

1. Theoretical Rigor: Complete mathematical proofs with refined technical treatment, particularly innovative handling of distribution dependence
2. Problem Importance: Fills important gap in statistical testing for McKean-Vlasov models
3. Methodological Innovation: Cleverly combines functional analysis, stochastic process theory, and high-frequency statistics
4. Practical Value: Provides implementable testing procedures with good asymptotic properties

Weaknesses

1. Lack of Numerical Verification: Pure theoretical work without simulation experiments
2. Strong Assumptions: Assumptions like functional derivatives may be difficult to verify in practice
3. Computational Challenges: Practical implementation may face computational complexity issues
4. Limited Application Guidance: Lacks specific guidance for practical applications

Impact

1. Academic Contribution: Pioneering significance in McKean-Vlasov statistical inference
2. Theoretical Value: Provides important theoretical foundation for subsequent research
3. Application Potential: Promising applications in financial risk management and systemic risk modeling

Applicable Scenarios

1. Financial Modeling: Verification of systemic risk and mean-field game models
2. Physical Systems: Modeling validation for large-scale interacting particle systems
3. Social Sciences: Statistical testing of collective behavior models
4. Engineering Applications: Validation of complex network system dynamics modeling

References

The paper cites 30 relevant references, primarily including:

• McKean-Vlasov theoretical foundations (Sznitman, 1991; Carmona & Delarue, 2018)
• Statistical estimation methods (Amorino et al., 2024; Belomestny et al., 2022)
• Classical SDE testing methods (Dette & Podolskij, 2008; Corradi & White, 1999)
• High-frequency statistics theory (Barndorff-Nielsen et al., 2006)

This paper makes important theoretical contributions to the field of statistical testing for McKean-Vlasov stochastic differential equations, providing a solid mathematical foundation for this emerging interdisciplinary area. While lacking numerical verification, the establishment of its theoretical framework provides a foundation for subsequent applied research.