Statistical analysis of child growth data
The study of child growth is complex. There are many clinical questions to answer but not necessarily the statistical methodology to deal with these questions. Human growth begins at conception and continues into adult life. In chapter 1 we discuss the characteristics of the growth process from conception to maturity and the purpose of growth monitoring. In chapter 2 we summarise the mathematical approaches to growth data. In chapter 3 we summarise the approaches that have been used to detect growth faltering. In this chapter we introduce the conditional gain Z-score. The data set analysed within this thesis is from the Newcastle growth and development study. In infancy we have routine weights of 3415 term infants. A sub-sample of these infants were followed-up at 7-9 years as part of a research study. These children belonged to three subgroups: cases were children that were defined as failing to thrive in infancy, controls were matched to cases and a 20% systematic sample. The school entry data of the sub-sample followed at 7-9 years were retrieved from school health records. In chapter 4 we carry out a preliminary analysis of the routine infancy weight Z-scores. The infancy data provided the opportunity to generate the correlation structure of routine weight Z-scores in infancy. In chapter 5 we develop a model for this correlation structure. In chapter 7 we explore patterns in the conditional weight gain Z-scores and also suggest some alternative criteria for identifying growth faltering in infancy. In chapters 6, 8 and 9 we analyse the anthropometric data obtained at follow-up and school entry. In childhood, the conditional gain Z-score is used to contrast height with mid-parental height and height at follow-up with height at school entry. The anthropometric data of the case and control children will be compared.