Statistical Modelling of the Incidence of Multiple Maternities

Project: Externally funded project

Project Details

Description

During the last years some projects have generated results of more general statistical interest. E.g. we have built a statistical model for recurrent twinning in the same family. The model is rather simple but the available data demand iterative numeric parameter estimation methods and therefore the convergence of such iterative processes had to be settled. (Fellman & Eriksson, 1990b) In other projects the shortcomings of the available data have also forced us to develop new statistical methods. In the study of twinning, the effect of the maternal age on the twinning rate is so important that a standardization of the data according to maternal age is necessary before the analysis of the effects of other factors can take place. However, the twinning data in different countries and for different time periods are of very divergent quality. Therefore we have developed new standardization methods applicable when other methods fail (Fellman and Eriksson, 1990b, 2002). Other scientists have also used our method with good results. A new still unsolved problem is that the effect of the maternal age on the prevalence of twinning shows strong regional and temporal variations (Fellman and Eriksson, 1993, 2004a). Furthermore, it has been observed strong seasonality of births and especially of multiple births and also the seasonality pattern shows both temporal and regional variations. We have generalised earlier presented ML models but we have also studied how trigonometric regression models (Fourier regression) can be used within this framework (Eriksson and Fellman, 2000a, Fellman & Eriksson, 1999, 2000b). The traditional statistical analyses of the twinning rates are usually based on rather simple assumptions. In order to study if these analyses are robust enough, we have compared them with more advanced models. Our results indicate that the simple methods are in general robust. Only the use of confidence intervals should be improved (Fellman and Eriksson, 2004b). During the 19th and the first half of the 20th century strong decreasing trends in the rates of the multiple maternities have taken place. This decrease in Sweden has eliminated the regional variation, which were strong during the 19th century. Now the different counties in Sweden show a common low level of the twinning rate (Fellman & Eriksson, 2005a). However, during the last decades of the 20th century a marked increase in the TWR can be noted in many populations. It cannot be explained by the slight increase in mean maternal age. The main cause of the recovery of the twinning rate seems to be the influence of artificial reproduction technologies and particularly the use of fertility-enhancing drugs (Fellman & Eriksson, 2006).
StatusActive
Effective start/end date01.01.200531.12.2020