讲座摘要:
In this talk, we consider partially linear single-index quantile regression with longitudinal data. By using Bayesian techniques, quasi-posterior distributions of the linear and single-index parameters are constructed based on a quasi-likelihood function. Under suitable assumptions, we derive asymptotic normality of posterior estimators of the parameters, and establish asymptotic relationship between the posterior estimators and their corresponding frequency estimators. Meanwhile, we use a stochastic search hierarchical model with spike-slab priors to perform variable selection and study consistency of the variable selection. Finite sample performance of the proposed methods is analyzed via simulation and real data too.