Análise bayesiana de dados funcionais com o uso de processo Gaussiano e metanálise: uma aplicação para a marcha humana

Abstract

The term "functional data" arised to accommodate situations in which each observation can be naturally interpreted as a function. These situations have become increasingly common with the availability of measuring instruments capable of recording large volumes of information with high frequency. In this context, a branch of statistics denominated functional data analysis emerged, in which appropriate methods of analysis were developed. An example of application of these techniques is in the study of human gait. The analysis of human movement is fundamental for understanding the normal movement and for proposing and evaluating preventive or rehabilitation programs. In this work, we will consider angular rotation data of the knee during the gait for a population of individuals without previous lesions with the aim of characterizing individual and population patterns. For this, we consider for each individual a Bayesian regression model with Gaussian process, and the conclusions was illustrated by the proposition of different methods for construction of functional predictive bands. To summarize the conclusions across the different individuals of the population, we propose to apply a Bayesian meta-analysis.