This project focuses on inference for stochastic processes, random fields, latent models, and regression problems with dependent errors. The common thread is the use of scale-wise information, especially through wavelet variance and related moment conditions, to estimate complex dependence structures in a way that remains computationally feasible for large data sets.
The work begins with robust wavelet-based inference for time series and random fields, where the objective is to estimate latent stochastic models without relying on fragile likelihood calculations. These ideas extend to spatial models and dependent data settings in which classical assumptions such as independence, correct parametric specification, or clean Gaussian errors are often unrealistic.
Several publications in this line study foundational questions: identifiability of latent models, robustness to contamination, computational speed, and the behavior of estimators for composite stochastic models. More recent work connects these ideas to large-scale regression with dependent errors, where dependence is not a nuisance to be ignored but a structure that must be modeled to obtain reliable inference.
Selected related publications include:
- Robust Inference for Random Fields and Latent Models
- Robust Inference for Time Series Models: A Wavelet-Based Framework
- Wavelet Variance for Random Fields: An M-Estimation Framework
- Fast and Robust Parametric Estimation for Time Series and Spatial Models
- On the Identifiability of Latent Models for Dependent Data
- Robust Two-Step Wavelet-Based Inference for Time Series Models
- Wavelet Variance Based Robust Estimation of Composite Stochastic Models
- Inference for Large Scale Regression Models With Dependent Errors
