Bayesian Nonparametric Vector Autoregressive Models via a Logit Stick-breaking Prior: an Application to Child Obesity

TL;DR

This paper introduces a novel Bayesian nonparametric vector autoregressive model with a logit stick-breaking prior, designed to identify and analyze childhood obesity patterns over time, demonstrating superior clustering and predictive capabilities.

Contribution

The paper develops a new Bayesian nonparametric model combining VAR and a dependent logit stick-breaking prior for clustering childhood obesity growth patterns.

Findings

The model effectively captures distinct obesity growth patterns.

Simulation studies show the model outperforms existing methods.

Application to real data reveals meaningful obesity clusters.

Abstract

Overweight and obesity in adults are known to be associated with risks of metabolic and cardiovascular diseases. Because obesity is an epidemic, increasingly affecting children, it is important to understand if this condition persists from early life to childhood and if different patterns of obesity growth can be detected. Our motivation starts from a study of obesity over time in children from South Eastern Asia. Our main focus is on clustering obesity patterns after adjusting for the effect of baseline information. Specifically, we consider a joint model for height and weight patterns taken every 6 months from birth. We propose a novel model that facilitates clustering by combining a vector autoregressive sampling model with a dependent logit stick-breaking prior. Simulation studies show the superiority of the model to capture patterns, compared to other alternatives. We apply the model to the motivating dataset, and discuss the main features of the detected clusters. We also compare alternative models with ours in terms of predictive performances.