Orateur
Gabriel Mastrilli
Description
Fitting parametric models to point process data is challenging because the likelihood is usually intractable. Standard approaches either replace it with a tractable surrogate, or fit summary statistics to their theoretical counterparts.
Whittle estimation is a middle ground: it builds a pseudo-likelihood from spectral statistics, doing inference in the frequency domain rather than directly on the point process. Recent work has shown this procedure leads to an estimator that is consistent and asymptotically normal.
In this talk, I will present ongoing work pairing this estimator with bootstrap confidence intervals, enabling uncertainty quantification in practice.
| Thématiques | Spatial point processes, Spectral inference |
|---|---|
| Mes travaux sont plutôt de la statistique ... | Théorique |
Auteur
Gabriel Mastrilli