Modelos para dados pré-teste/pós-teste no contexto clássico e bayesiano

Abstract

In many practical situations, the interest lies in modeling data from studies with repeated measures. To this end, it is important to develop techniques for shaping, setting and assess the quality of the data model used for these criteria. Much of the effort employed in data analysis with repeated measures is related to the dependence structure modeling intra-experimental units. Within the class of studies with repeated measures, it is worth mentioning the studies with data type pre-teste/pos-teste. Statistical models for data pre-teste/pos-teste been considered by several authors. These models are characterized as longitudinal studies whose observations are taken before and after any type of intervention, treatment, for example, to study their effects over time. There are different ways in which experiments with data pre-and post-test are conducted and specific models for this approach are already very devoted. From the use of two datasets previously analyzed (Sef, 1999, Singer et al., 1988), given the existence of influential observations and leverage in the context of linear mixed models, as highlighted in Noble and Singer (2007, 2011), propose to introduce the use of mixed linear models, Bayesian and the classical context, as one of the most used tools to model data type pre-teste/pos-teste, in order to obtain more robust models for prediction.