Comparison of two methods for transforming height and weight to normality

I assessed the success of the Box-Cox power transformation and the shifted log-transformation for transforming length/height and weight to Normality. Retrospectively collected data for full-term, healthy 1143 boys and 1162 girls amounted to over 45000 recordings. Of the children, 48%, born during 1959-1961, were followed from birth to age 19.0, and the rest, born during 1969-1971, until age 12.5. The individual growth curves were smoothed in order to obtain reliable interpolated estimates of measurements at fixed ages and to avoid age grouping. Height is fairly Normal, even during puberty, and weight can be transformed close to Normality, the Box-Cox transformation being preferable to the shifted log-transformation. Occasionally remaining leptokurtosis in the transformed weight can apparently be disregarded when focusing on the extreme (1st, 2.5th, 97.5th and 99th) centiles. The 95% confidence intervals for the two transformation parameters are wide, and may even explode for the shift parameter. The shapes, although not the levels, of the curves joining the estimates are similar for boys and girls, showing a 2-year difference in maturation. Improved screening for abnormal growth, based on the assumption of Normality, after a transformation, seems possible.