Bayesian Semiparametric Estimation with Nonignorable Nonresponse
share
Statistical inference with nonresponse is quite challenging, especially when the response mechanism is nonignorable. Although existing methods often require correct model specifications for response models, the models cannot be verified based on the observed data and misspecification of the response models can lead to a seriously biased inference. To overcome this limitation, we develop an effective Bayesian semiparametric method for the response mechanism using logistic penalized spline methods. Using Polya-gamma data augmentation, we developed an efficient posterior computation algorithm via Markov Chain Monte Carlo. The performance of the proposed method is demonstrated in simulation studies and an application to a longitudinal data.
READ FULL TEXT
