Introduction

Anthropometric measurements are used in a wide number of contexts in human biology. For instance, twin researchers compare monozygotic and dizygotic twins and find very high heritabilities for height and weight (Bouchard et al., 1990). Others use height-for-age or weight-for-height of children to show the extent of under nutrition in a study area (Waterlow et al., 1977). In adults low body mass index is thought by many to be a useful predictor of increased morbidity, reduced work capacity and increased risk of mortality (Ferro-Luzzi et al., 1992).

The aim of this chapter is to discuss some of the main statistical issues and approaches when analysing cross-sectional anthropometric characters. The paper is divided into three sections. The first section deals with sample design and sample size determination; the second section concentrates on analyses of continuous characters and covers testing of normality, use of analysis of variance with main effects and covariates, testing for curvilinearity and the use of regression. The final section addresses the issue of the use of statistics for predictive purposes. It reviews the use of relative risk and odds ratio, sensitivity and specificity, logistic regression, discriminant analysis and modelling using multiple regression. Examples of many of the statistical tests and interpretation of results are provided.

Sample design and sample size determination

Study design

The term study design incorporates the type of research design, definition of the study sample, determination of sample size, method of treatment allocation (if applicable) and means by which the sample is obtained (random, cluster, quota, sequential, etc.). There are only two types of research design: either experimental or observational. In experimental studies the intervention is under the control of the researcher.