Statistical
applications of over-fitting due to trimmings

Pedro Alvarez-Esteban

Universidad de Valladolid, Spain

Roughly speaking, the underlying principle that we will explore, says
that if two random samples of the same probability distribution are
partially trimmed to make them as similar as possible, then you should
be able to distinguish these pair of trimmed samples from any other
pair of non-trimmed samples of the same sizes.
In this talk, our goal will be to provide sound and empirical evidence
of this aﬃrmation and design a general bootstrap procedure for
comparison of two samples or one sample and a given distribution. This
statistical procedure should be also useful in other frameworks of
model validation.
As an added value of the principle, we provide an appealing methodology
to analyze, from a non-parametrical point of view, if we can assume
that k samples arise from essentially identical underlying structures.