Finding better ways to chart growth over time

Authors: JO, CHAN-HEE - ACHRI-DAC; GOSSETT, JEFF - ACHRI-DAC; SIMPSON, PIPPA - ACHRI-DAC; Bogle, Margaret

Submitted to: International Society for Behavioral Nutrition and Physical Activity
Publication Type: Abstract Only
Publication Acceptance Date: 4/20/2007
Publication Date: 6/16/2007
Citation: Jo, C., Gossett, J., Simpson, P., Bogle, M.L. 2007. Finding better ways to chart growth over time [abstract]. Proceedings of the International Society for Behavioral Nutrition and Physical Activity. p. 237.

Interpretive Summary:

Technical Abstract: Our purpose was to show how longitudinal data, especially anthropometric data such as height, weight, and BMI, may be better modeled using non-parametric methods. In many nutrition and physical exercise interventions, subjects are observed and/or measured on multiple occasions. To account for the longitudinal nature of the data, a mixed models analysis is commonly used. It is typical to make comparisons between dose or treatment groups, possibly controlling for demographic variables. When the dose response curve is non-linear, as growth typically is, it can be difficult to find an appropriate curve, and model misspecification is common, leading to less than optimal comparison. Many interventions cover relatively short time periods, and demonstrating effects using means is not always satisfactory. We will present an alternative to non-linear mixed models that avoids the problem of specifying a parametric model. We demonstrate the use of a semi-parametric regression, namely splines, in a mixed model framework, using easily available software. This use allows for nonlinear fitting with a covariance structure for the correlated observations. We illustrate its use on longitudinal BMI data with various time points, like the US National Longitudinal Youth Survey. Regression splines may give a more sensitive approach to evaluating changes in outcomes for nutrition and exercise interventions.