2023 Fiscal Year Research-status Report
Sparse statistical approach for multivariate modelling
| Project/Area Number |
22K13377
|
|---|
| Research Institution | Osaka University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2022-04-01 – 2025-03-31
|
| Keywords | Copula / Factor model / High dimension / Sparsity |
|---|
| Outline of Annual Research Achievements |
The paper "Sparse M-estimators in semi-parametric copula models", co-authored with Prof Fermanian, has been accepted for publication at Bernoulli in 2023 and is forthcoming in 2024. The paper answered the issues inherent to copula models: pseudo-observations; unbounded copula-based objective functions; explosive number of parameters. We specified a suitable penalized M-estimator framework for copulas and derived the asymptotic properties.
The paper "Sparse factor models of high dimension", co-authored with Prof Terada, is currently submitted at an econometrics journal: we devised a sparsity-based estimation framework for the factor loading matrix taking into account the rotational indeterminacy and derived the asymptotic properties.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
The research is moving forward: one paper published in Bernoulli; one paper currently submitted at an econometrics journals; one research project on stochastic volatility models is about to be completed.
|
| Strategy for Future Research Activity |
The project "Factor Multivariate Stochastic Volatility Models", co-authored with Prof Asai, is about to be completed and will be submitted at an econometrics journal. The key idea is to integrate factors in the Multivariate Stochastic Volatility (MSV) model. We propose to estimate the latent factors using the estimators of the factor decomposition and then specify a multivariate state space representation of the latent volatility of the factors (not the observed random vector, which can be high-dimensional). Theoretical analysis of the proposed method: asymptotic properties derived under moment conditions.
The replication package will be made publicly available for the sake of transparancy.
|
