Bayesian geoadditive modelling of breastfeeding initiation in Nigeria
利用1999年尼日利亚人口与健康调查数据，通过贝叶斯地理加性模型分析母亲年龄和地理位置对产后首次母乳喂养时间的影响，为公共卫生政策提供参考。
Abstract A study into the geographical variability of timing of initial child breastfeeding after birth was carried out with the data set from the 1999 Nigeria Demographic and Health Survey. The effect of the metrical covariate of the mother's age at birth was assumed to be nonlinear and estimated nonparametrically. Other categorical covariates are estimated in the usual parametric form. Within a Bayesian context, appropriate priors are assigned for the geographical location, vector of the unknown (nonlinear) smooth functions and a further vector of the fixed effect parameters. For instance, a Markov random field prior is assumed on the spatial effects. Inferences are based on Markov chain Monte Carlo techniques while Bayesian model diagnostics are based on the deviance information criteria. Copyright © 2004 John Wiley & Sons, Ltd.