Document Type



This paper discusses the sources of error in the ages of intersection of growth curves of boys and girls. First, the impact of the sample design (cross-sectional or longitudinal) on the interpretation of observed points of intersection is illustrated by fictional examples. It is further shown that errors in estimated ages of intersection are related both to sample errors (as compared to population growth curves) and to the angle of intersection. To calculate such errors a stochastic computer simulation technique is introduced. Calculations are made on the basis of data of the Nymegen Growth Study. For 12 anthropometric quantities the first age of intersection and its error are estimated. Results show that for four quantities (bicondylar diameters of femur and humerus, head circumference, leg length) there is no intersection; boys have greater mean values than girls at all ages. For three quantities (weight, sitting-height, sitting-ratio) girls’ mean curves cross those of boys with standard errors of 1.0, 0.7, and 0.9 years, respectively. These errors are relatively large as com­pared with the duration of the female ascendancy, with possible shifts over time within one population, or with possible differences between populations. For the remaining five quantities (height, biacromial and biiliocristal diameters, circumferences of arm and fore-arm) it proved to be doubtful to take intersection into consideration, according to the analysis technique used. These results indicate that several serious methodological pitfalls may exist in the study of female ascendancy.