[Using parametric and non-parametric methods to estimate the pattern of the age at natural menopause based on prevalence data]

ABSTRACT

Various studies have reported different ages of menopause in Iran, from 46 to 51 years. This broad range of age at menopause cannot be attributed solely to the difference between women of different regions, but could also be related to methodological problems. The model of menopausal age could be estimated by a cohort study or a cross-sectional survey and each method has advantages and disadvantages. In cross-sectional studies, the target is only the status of menstruation for each woman, such that the prevalence of menopause can be calculated in each age group. Therefore, the pattern of the age at menopause cannot be estimated using the prevalence data. For estimating the patterns of the age at natural menopause and its characteristics, parametric and non-parametric methods were utilized and the data from the survey "Estimating the pattern of the age at natural menopause and its effective factors in Semnan" was used to illustrate these two methods. In that survey, during the year 2007, 3545 urban and rural women were randomly selected and their status in terms of menopause was asked, then by excluding abnormal menopause, 3404 had been selected for the study. The pattern of menopause age was obtained with the help of generalized linear models [GLM] with parametric base and compared with non-parametric method in which the probability distribution is estimated empirically. By applying and comparing the two parametric and non-parametric methods the method that provides better natural menopause age model was clarified. Although there were not many differences between the estimated values in the different models, but the log-log model provided a more accurate estimation than the two-parametric, logistic and probit models and its results were similar to the non-parametric model. It seems, logistic model is more appropriate than other models to estimate the pattern of the age at natural menopause, although further research is needed