Estimating Park-Based Moderate to Vigorous Physical Activities by a Bayesian Infinite Server Queue Model

• Activity Number: 640
• Type: Topic Contributed
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Bing Han*+
• Companies: RAND Corporation
• Keywords: Physical activity ; Bayesian analysis ; Infinite server queue ; continuous time Markov process ; semiparametric ; Neighborhood park

Abstract

Neighborhood park facilities are important resources for moderate to vigorous physical activity (MVPA) for urban residents. Large-scale measurements of park-based MVPA are often based on repeated snapshot observations (e.g., counts of park users engaging in MVPA at observation times) and surveys (e.g., self-reported frequency and duration of park-based MVPA). This study applies a nonhomogeneous infinite server queue model to estimate park-based MVPA, in which new MVPA starts according to a non-homogeneous Poisson process and the duration of individual sessions follows an exponential distribution. We propose a semi-parametric Bayesian approach to fitting the probability model by using both linear effects of environmental and demographic predictors and a flexible temporal trajectory. The potential biased self-report data are used to derive a subjective prior distribution for some important parameters. The model likelihood is based on the objective snapshot observation data. We develop an efficient MCMC algorithm for calculation. We analyze the MVPA and park use of two Los Angeles neighborhood parks.