Adaptive and sparse estimation in the functional linear model
The aim of functional data statistics is to study data that can be represented as curves (temperature, electricity consumption,...).
The aim of this talk is to present recent works on estimation in the functional linear model which is a linear model whose covariates are functional data.
We first study the case where there is only one functional covariate and propose a projection estimator based on PCA. We define a data-driven criterion for selecting the dimension of the projection space and show that the estimator selected achieves the optimal convergence rate in the minimax sense.
In a second part, we will focus on the multivariate functional linear model where the covariates are multiple and can be of different nature and we will study the theoretical properties of a variant of Lasso.
The method will be motivated and illustrated on simulated and real data sets.
Practical information
- Informed public
- Free
Organizer
- Victor Panaretos
Contact
- Maroussia Schaffner