25. november 2015

Giueppe Cavaliere's study is accepted by Annals of Statistics

A long standing open issue in econometrics and statistics is how to make inference in time series which feature infinite variance [IV] innovations. This framework is particularly challenging because under IV the asymptotic distributions of estimators are in general non-standard and the bootstrap tends not to be robust to infinite second order moments. In this paper the authors fill an important gap in the literature by showing that two bootstrap methods (based on permutation and wild bootstrap algorithms) are consistent and can be applied under very general assumptions.

Giueppe Cavaliere's study is entitled " Sieve-based inference for infinite-variance linear processes "  and is a joint paper with Iliyan Georgiev and Robert Taylor