WeGrowHealthR: Reproducible Analysis and Modelling of Child Growth Data in R

Abstract

Abstract

Child growth studies combine international growth references, longitudinal models and prediction tools, and errors tend to arise where these pieces meet. WeGrowHealthR is an R package that brings data checking, World Health Organization and Centers for Disease Control and Prevention z-scores, growth curve models, quantile methods, growth phenotypes, multistate models and prediction benchmarks into one consistent interface. A data contract object standardises longitudinal growth data. Report objects collect the tables, figures, models and sample flow of an analysis together with the success or failure of every model, and run modes switch a whole analysis between quick development settings and full publication settings. Fourteen study flows each reproduce a complete analysis in one call. Parallel computing gives every task its own random stream, so results are identical with any number of workers. A simulation engine produces cohorts with known truth, which the package also uses to test its own estimators. This article describes the design, shows a typical workflow and a full analysis of the pubertal growth spurt on simulated data, and reports agreement with the WHO packages anthro and anthroplus, recovery of known quantities over repeated simulated cohorts, and computing times. The package does not introduce new statistical methods and makes no claim of clinical validity.