Influences on Menarcheal Age in Girls in a Welsh College

Of the methods used to obtain estimates of the mean and dispersion of age at menarche in a population, the potential accuracy of two are generally agreed: the prospective method, in which individuals are followed in longitudinal studies throughout the pubertal period, and the status quo method, in which probits or logits are fitted to data from Yes/No questions to girls of known age in the range 9 to 17 years. These two procedures avoid many of the errors and biases that may arise from cross-sectional retrospective studies in which girls are asked to recollect their age when they first started to menstruate. Such errors may arise from the difficulty of exact recollection; this is particular ly so in the case of older women but may also occur at younger ages, for De Wijn (1966) showed that the percentage of teenage girls unable to recollect the exact month increases with the age at which they are asked; but errors of recollection do not always occur, as was shown by Livson and McNeill (1962), who compared recorded age with that remembered 17 years later and found close individual and very close mean concordance. Much depends on the intelligence of the girl and on the clarity of definition of the event. There may be bias if there are any girls in the sample who have not yet menstruated or who give deliberately false answers to emulate their friends; for example, Kark (1943) showed that girls of 18 to 20 gave an average age remembered as 15 to 16 years, whereas those of 15 to 17 gave an average age of 13 to 14 years; Wilson and Suther land (1949), comparing two groups of children in Oxford and in Oxfordshire, snowed that the difference in their respective remembered ages (13-4 and 13*09 years respectively), was entirely due to the inclusion of more pre-pubertal girls in the latter sample, for probit analysis showed an identical