A0369
Title: Bayesian quantile scalar on image regression with non-ignorable non-response
Authors: Qi Yang - Department of Statistics, the Chinese University of Hong Kong (Hong Kong) [presenting]
Xinyuan Song - Chinese University of Hong Kong (Hong Kong)
Abstract: Exploiting imaging information is usually difficult because imaging data have ultrahigh dimensions. Functional principal component analysis can be used to reduce the dimensionality of imaging data, such that the reduced imaging information can be incorporated into conventional regression as predictors. However, a mean regression analysis generally does not provide a comprehensive understanding of the relationships between covariates and responses of interest. Instead, quantile regression enables the model to assess such relationships at different quantiles and simultaneously provides robust estimation results regardless of response distributions. We consider a quantile scalar on image regression model in the presence of nonignorable missing data. The proposed model includes ultrahigh dimensional imaging data as predictors to examine the effects of medical images and other covariates on the scalar response of interest. We develop a Bayesian approach with Markov chain Monte Carlo algorithms to conduct statistical inference. Simulation studies demonstrate that the proposed method performs satisfactorily. An application to a study of the Alzheimer's Disease Neuroimaging Initiative dataset is presented.
