Nonparametric Distribution Fitting of Age at First Marriage of Nepalese Women through Scatter Plot Smoother

Abstract

Age at first marriage is one of the important aspects for studying fertility behavior of women since it is believed that the number of children ever born is inversely related to age at first marriage. An attempt has been made to fit the Nepal Demographic and Health Survey (NDHS)-2006 data for age at first marriage of Nepalese women. The lack of fitting of some parametric distributions like lognormal and weibull was detected when applied to population census data of Nepal-2001 and the NDHS data-2006. Even though there is possibility of fitting more complex theoretical distributions, this highlighted the need to explore for alternative methods. In the process, it was identified that nonparametric smoothing is perhaps one of the viable solutions to this problem apart from fitting more sophisticated parametric distributions. Consequently, two nonparametric smoothers were examined for their suitability for distributional fitting, namely locally weighted regression smoother (LOESS) and cubic smoothing splines. Finally, cubic smoothing splines was chosen and fitted to NDHS data for Nepal and also separately for different ecological belts of Nepal since it fitted better than LOESS.

Key words: cubic smoothing splines; locally weighted regression smoother; Nepal demographic and health survey; nonparametric smoothing; theoretical probability distributions

DOI: 10.3126/njst.v9i0.3184

Nepal Journal of Science and Technology 9 (2008) 179-186