Some methods of temporal statistics are presented and proposed for investigating the date of birth as a disease predictor. A subject's birthday is proposed to be used as a continuous variable with a circular distribution, a special type of interval scale without a true zero point. Three types of endpoints are then considered: a dichotomous endpoint; a continuous endpoint; and time-to-event. A study of otitis media is used for illustration. We found, for example, that children born in late winter to early spring tend to have higher cord blood pneumococcal antibody concentration and lower risk of disease as compared to those born in summer to early fall perhaps due largely to exposure to indoor pollution by pregnant mothers.
Copyright 2003 John Wiley & Sons, Ltd.