Author/Authors :
Mohd Din, Siti Haslinda Erasmus University - Medical Centre - Department of Biostatistics, Netherlands , Mohd Din, Siti Haslinda Department of Statistics, Malaysia , Koon, Poh Bee Universiti Kebangsaan Malaysia - School of Healthcare Sciences, Faculty of Health Sciences, Malaysia , Noor, Mohd Ismail MARA University of Technology - Faculty of Health Sciences, - Department of Nutrition and Dietetics, Malaysia , Henry, Christiani Jeya K. Oxford Brookes University - Functional Food Centre, UK , Henry, Christiani Jeya K. Singapore Institute for Clinical Sciences - Clinical Nutrition Research Centre, Singapore , Lesaffre, Emmanuel L-Biostat, KU Leuven, Belgium , Lesaffre, Emmanuel Erasmus University Medical Centre - Department of Biostatistics, Netherlands
Abstract :
A basal metabolic rate(BMR) that is too low is an indicator of a poor physical condition, and could be one of the reasons for overweight. Measuring BMR though, is a time-consuming exercise, and there has long been interest in developing statistical models to predict BMR from demographic and anthropometric measurements. Poh et al. [1] developed ordinary linear regression models on a cohort of 139 Malaysian children measured three years bi-annually. However, since each child contributed six times to the total data set, these models ignore the correlated nature of the data. We re-analyzed these data using correlated linear models. We show that our approach taking correlation into account is important to establish important covariates, but does not improve prediction.
