Title: Identification, estimation, and semiparametric efficiency of nonignorable missing data
Authors: Wang Miao - Peking University (China) [presenting]
Abstract: Identification and estimation with an outcome missing not at random will be discussed. We note that without extra assumptions, even fully parametric models are not identified. We then study identification strategies based on auxiliary variables such as an instrumental variable or a shadow variable. An instrumental variable impacts the missingness, but not the outcome of interest, and in contrast, a shadow variable is correlated with the outcome, but independent of the missingness after conditioning on the outcome. We describe general conditions for nonparametric identification of the full data law. We describe semiparametric estimation methods, and we characterize the semiparametric efficiency bound for the class of doubly robust regular and asymptotically linear estimators.