Longitudinal and Cross-sectional Reference Curves of Gross Motor Function in Children

Abstract

<p>Reference curves are the most popular tool to monitor the time-related growth in children. Both cross-sectional and longitudinal studies are widely used to collect the reference samples. The methods used for constructing the reference curves and the interpretation of the curves for longitudinal studies should be different from those for the cross-sectional studies. However misunderstanding in constructing and interpreting the reference curves for the cross-sectional and longitudinal studies is common, especially the concerning of the effect of regression to the mean in the longitudinal studies. The LMS models of Cole and Green[1,2] using penalized likelihood are considered to be the most powerful methods to construct the reference curves for both cross-sectional and longitudinal studies. This thesis focuses on the comparison of the conditional LMS regression approach for drawing the median conditional centiles for longitudinal data to the conventional LMS model for constructing the distance centiles for cross-sectional data. It describes the different interpretations of the two approaches. The application of the two methods to a study of Gross Motor Function is investigated in detail to illustrate the difference between them.</p>