Research projects

Project CEMAPRE internal

TitleBayesian ANOVA
ParticipantsGonzalo Garcia-Donato, Rui Paulo (Principal Investigator)
SummaryAnalysis of variance (ANOVA) models are linear models where the design matrix is not full-rank. Rank
deficiency has very well-known important consequences at the estimation and testing levels. For
example, only estimable functions of the parameters can be estimated, which implies that the effects
of the different levels of the factor on the response cannot be estimated. The usual solution to
rank deficiency is based on the notion of full-rank reparameterization of a model. We will address
the problem of objective prior specification and model selection in this context in an effort to
find alternatives approaches to analyzing ANOVA models that do not require a full-rank
reparameterization stage.