A1511
Title: Robustified Gaussian quasi-BIC for volatility
Authors: Shoichi Eguchi - Osaka Institute of Technology (Japan) [presenting]
Hiroki Masuda - University of Tokyo (Japan)
Abstract: There have been several studies on model selection for stochastic differential equations and robust model selection; for example, the BIC-type information criterion for locally asymptotically quadratic models and AIC- and BIC-type information criteria based on density-power divergence. In this talk, robust model comparison is considered in a class of non-ergodic continuous volatility regression models contaminated by finite-activity jumps. Using density-power weighting and Holder-inequality-based normalization of the conventional Gaussian quasi-likelihood function, two Schwarz-type statistics are proposed.
