Work

Integrated likelihood for the treatment of nuisance parameters

Year: 2012
Type: dissertation
Abstract: In literature, several tools have been proposed to make inference about a parameter of interest ψ in presence of nuisance parameters. Among these, the integrated likelihood seems to gain popularity. Commonly used in Bayesian inference, the integrated likelihood has been recently studied also under the frequentist paradigm. We contribute to this analysis studying first its properties in presence of many nuisance parameters, and in particular in situations when the number of nuisance parameters increases with the sample size. In this setting, indeed, the usual inferential tools, based on the profile likelihood, may perform poorly, and the use of the integrated likelihood can be an alternative to higher order methods. In particular, we focus on the asymptotic behaviour of the signed square root integrated likelihood ratio statistic, studied in a two-index asymptotics setting, in which both the sample size and the dimension of the nuisance parameter increase to infinity. As a second topic of the thesis, when there is a sufficient statistic for the nuisance parameter, we study conditions of equivalence of integrated and conditional likelihoods. Finally, some efforts are done to study the effect of the presence of nuisance parameters on pairwise likelihood and on the related score function. A correction useful to reduce the profile pairwise score bias is presented.
Language: English
Open Access status: green
OpenAlex ID: W77281370
Cites: 49
Cited by: 1
Related to: 9