Curve matching: a data-driven technique to improve individual prediction of childhood growth

Abstract

Longitudinal growth data are valuable for predicting and interpreting future growth of individual children. This note explores the idea of 'curve matching', a new technique to improve prediction of future growth of an individual child. The key idea is to find existing children in existing databases that are similar to the current child. The growth patterns of the matched children suggest how the current child might evolve in the future. This paper describes the various conceptual and practical issues that need to be addressed before the idea can take off. A demo implementation is available at http://vps.stefvanbuuren.nl:3838/frisodemo/. © 2014 S. Karger AG, Basel.