A0482
Title: Profile quasi-likelihood inference for SDE with mixed effects
Authors: Hiroki Masuda - University of Tokyo (Japan) [presenting]
Maud Delattre - INRAE (France)
Abstract: Mixed-effects models play a pivotal role across various scientific domains, facilitating the precise analysis of repeated observations among individuals. Recent advancements propose random-effect modeling based on the generalized hyperbolic distribution to better accommodate variability. Utilizing normal variance-mean mixture-type random effects is considered in a class of stochastic differential equations (SDE) with mixed effects. The statistical framework allows for the exploration of a wider class of mixed-effects diffusion models compared to previous literature. A novel parameter estimation method is proposed, and theoretical insights are provided into the asymptotic behavior of the estimators. The estimation method diverges from the quasi-likelihood approach, offering a more accessible numerical procedure while sacrificing some efficiency compared to maximum likelihood estimation. This trade-off ensures stability and ease of implementation in high-frequency frameworks.
