[A simple statistical method for achieving reliability in anthropometric measurements].

Abstract

Most anthropometric studies fail to comment on the reliability of the measures. The authors assume that the complexity of the classical analysis of variance approach is one of the reasons for this. They therefore describe a simpler, yet scientifically sound method which assesses three sources of measurement error: intra and inter-observer as well as instrument errors. The measure of upper-arm-circumference of pregnant women, which has been proposed to be a predictor of birth-weight, is used as an example. Reliability was assessed by calculating agreement of a series of pairs of measurements. Inter and intra-observer-errors were found to be of the same size. Both types of human errors were statistically significantly greater than the error arising from the use of two different types of tapes (instrument error). It is hoped that other authors find this approach useful so that information on reliability in anthropometric studies becomes the rule rather the exception.