Key research themes
1. How can singular and truncated moment matrix properties determine the existence and uniqueness of representing measures in moment problems?
This theme focuses on the structural characteristics of moment matrices, especially singularity and flat extension properties, and their implications for solving truncated moment problems. It emphasizes how positivity, recursive generation, and rank conditions of associated moment matrices govern the existence, uniqueness, and minimal atomicity of representing measures. Insights here provide foundational criteria to address representability in complex moment problems and their minimal quadrature rules.
2. What numerical and analytical techniques can be used to compute or approximate moment integrals and heat kernels for stochastic and physical processes?
This theme encompasses the development of computational methods and analytical approximations for moment integrals, heat kernels, and related kernels arising in stochastic processes and physical models such as diffusion, degradation, and quantum systems. It covers methods including moment recursion relations, Bell polynomial expansions, discrete Green’s theorem-based algorithms, and operator-theoretic expansions, aimed at efficient, precise, or series-based evaluation of complex integrals or probability functions without direct integration, facilitating applications in option pricing, stochastic modeling, and image analysis.
3. How can moment selection methods in econometrics be adapted to handle mixed identification strength in moment condition models?
This research area investigates statistical model selection techniques, particularly generalized method of moments (GMM) estimators, under the challenging scenario where moment conditions exhibit varying identification strengths, ranging from strong to weak to semi-weak. It addresses the inconsistency of classical selection criteria that fail to adjust for the parameter convergence rates aligned with identification degrees, proposing novel penalization and entropy-based criteria that incorporate estimation rates and asymptotic distribution properties. These advancements enable consistent and efficient selection of relevant moments, crucial for reliable inference in econometric models with heterogeneous identification patterns.