Uniform inference for Small Area Parameter

Título: Uniform inference for Small Area Parameter

 Ponente: Stefan Sperlich (Universidad de Ginebra, Instituto de Economía y Econometría, Suiza)

Fecha: Jueves 24 de mayo de 2018 a las 12:00 horas

Lugar: Sala de Seminarios, Instituto Universitario de Investigación CIO, Edificio Torretamarit, Universidad Miguel Hernández (Campus de Elche)

Resumen:

Today, SAE is a common tool used world-wide by Statistical offices for addressing the need of disaggregated information. Interval estimates can either be extremely wide if not model-based, or only refer to marginal (ie unconditional) distributions. That is, when speaking of a 95% confidence interval, for 5% of the considered areas, the intervals do not contain the true parameter. This is a delicate default if political decisions based on them, and prohibits the comparing areas based on those estimates.  In this work, construction of uniform prediction intervals (or simultaneous confidence sets) for small area parameter in linear mixed models is introduced. We consider three frameworks to develop simultaneous intervals: analytical, numerical and bootstrap approximation. Proofs of the consistency as well as the asymptotic coverage probability of the bootstrap intervals are provided. Our proposal is accompanied by simulation experiments and data examples.