Box-Jenkins modeling of chronological series of twin births and single births.

A time-series analysis by Box-Jenkins modeling of the monthly observations of twin births and singleton births was attempted. The study is population based. The single-birth series has been matched to the twin-birth series for maternal birth cohort, parity of the twin birth, and residence of the parents. Three-hundred forty-one twin pairs and 340 singleton births were grouped into chronological series spanning 72 months of observation. Univariate and bivariate time-series analyses were used. Box-Jenkins modeling shows that both series share the same basic demographic and reproductive risk factors, such as ovulation and conception rates, abortions, sexual intercourse, and the less well-known random events that determine the time of delivery. These factors are randomly related to both birth processes but exert a more important influence on twin births. Beyond these crude features, twin- and single-birth series differ by the frequency of their seasonal cycle. Twin births show low-frequency variations because of a rather limited number of factors that impinge markedly on the twinning process. Single births are subjected to higher frequency variations that can be ascribed to a greater number of structuring factors, the action of which is less pervasive. Only the January peak is common to both types of births. Aside from the common features, the bivariate analysis shows that both series of births are independent, neither being related to the other. It is concluded that the subtle structuring of both processes is largely distinct. The large fraction of the unaccounted variation of each birth process (about 70%) suggests that most of the variation is due to as yet unidentified factors. The search for environmental and geocosmic factors at the origin of conception and confinement should be considered in future undertakings.