|
Abstract |
We consider estimating mean functionals with nonignorable missing. Traditional methods handling nonignorable missingness usually entail parametric modeling of the missing mechanism. Kim and Yu (2011, JASA) proposed a semiparametric exponential tilting model for the same end. But their method requires an independent survey or validation sample to achieve identification. In this study, we consider the sensitivity of estimators of the mean functionals within Kim and Yu's exponential tilting model without assuming the availability of an independent survey or validation sample. Our proposed method can be viewed as a robust version of Kim and Yu's. |
