شماره ركورد كنفرانس :
4109
عنوان مقاله :
Bayesian nonparametric weighted sampling inference using Dirichlet process mixture model
پديدآورندگان :
Rahnamay Kordasiabi S Department of Statistics, Razi University, Kermanshah, Iran , Khazaei S Department of Statistics, Razi University, Kermanshah, Iran
كليدواژه :
Sample survey inference , Survey weighting , Poststratification , Dirichlet process mixture model , Bayesian nonparametric regression
عنوان كنفرانس :
يازدهمين سمينار ملي احتمال و فرآيندهاي تصادفي
چكيده فارسي :
We study Bayeian inference for the population mean in a design-based survey context. The nonsampled units values are not required for the usual Horvitz-Thompson estimator. The main focus of the paper is to incorporate information available from external sources to improve estimation and inference for finite population quantities. In Bayesian method for the finite population inference, Si et al. (2015) showed that including the weights of the nonsampled units in the population as predictors in a nonparametric Gaussian process regression can result in improved point estimator of the finite population mean. We propose a Bayesian nonparametric framework with Dirichlet process mixture model that is incorporating the wieghts as available information to imput the nonsampled units.
