Flexible Bayesian spatial modeling for unknown missing data mechanism in survey analysis: An application to the Chinese general society survey
Guanyu Hu, Ming-Hui Chen, Zhihua Ma
The Annals of Applied Statistics · 2025 · Volume 19 · Issue 2
Abstract
Social survey data are often collected from different survey centers in different regions. In some circumstances the response variables are completely observed while the covariates have missing values. In addition, the missing data patterns will vary in different geographical locations. In this article an ordinal response regression model with a log linear model for the cutpointsis extended to fit the spatial ordinal response data with missing covariates within a Bayesian framework. A signed-spike-and-slab prior is developed to learn the heterogeneity of the missing data mechanisms among different spatial locations. The properties of the proposed models are examined, and a Markov chain Monte Carlo sampling algorithm is used to sample from the posterior distribution. Extensive simulation studies are carried out to examine the empirical performance of the proposed methods. We further apply the proposed methodology to analyze a real dataset from a Chinese General Society Survey.
Keywords
conditional autoregression, nonignorable missingness, personal income, spatial heterogeneity
Enable JavaScript for citation tools and additional paper interactions.