Dombry, C. and Esstafa, Y.
The vanishing learning rate asymptotic for linear $L^2$-boosting, ESAIM PS, forthcoming.
Efficient
inference
for large and
high-frequency
data
The statistical decision theory deals with the problem of constructing an optimal decision for a given statistical experiment.
The statistical decision theory deals with the problem of constructing an optimal decision for a given statistical experiment.
The theory of statistical experiments deals with the convergence of statistical experiments and the construction of asymptotical optimal decision.
The theory of statistical experiments deals with the convergence of statistical experiments
and the construction of asymptotical optimal decision.
Since the convergence (in a reasonable sense) on the initial sequence of statistical experiments cannot be expected…
Since the convergence (in a reasonable sense)
on the initial sequence of statistical experiments cannot be expected…
…localized statistical experiments are built (mimicking the centering and renormalization in the central limit theorem).
…localized statistical experiments are built
(mimicking the centering and renormalization in the central limit theorem).
The convergence of the sequence of localized statistical experiment to a “simple” canonical experiment for which the optimal decision can be defined…
The convergence of the sequence of localized statistical experiment
to a “simple” canonical experiment for which the optimal decision can be defined…
…allows to define the optimal decision in the localized statistical experiments for sufficiently large size of the sample.
…allows to define the optimal decision in the localized statistical experiments
for sufficiently large size of the sample.
Moreover a “global” optimal decision can be built in the initial corresponding statistical experiment.
Moreover a “global” optimal decision can be built
in the initial corresponding statistical experiment.
The project aims to improve the knowledge on asymptotic efficiency and to provide new and innovative efficient estimators and testing procedure for large and high-frequency dataset encountered in real applications.
The project aims to improve the knowledge on asymptotic efficiency
and to provide new and innovative efficient estimators and testing procedure
for large and high-frequency dataset encountered in real applications.
Sponsors:
[A]Toggle Animation
[I]Invert colors
[G]Toggle Guides