Model Checking for a General Linear Model with Nonignorable Missing Covariates

In this paper, we investigate the model checking problem for a general linear model with nonignorable missing covariates. We show that, without any parametric model assumption for the response probability, the least squares method yields consistent estimators for the linear model even if only the complete data are applied. This makes it feasible to propose two testing procedures for the corresponding model checking problem： a score type lack-of-fit test and a test based on the empirical process. The asymptotic properties of the test statistics are investigated. Both tests are shown to have asymptotic power 1 for local alternatives converging to the null at the rate n-r, 0 ≤ r 〈 1/2. Simulation results show that both tests perform satisfactorily.

Variable screening with missing covariates: a discussion of ‘statistical inferencefor nonignorable missing data problems: a selective review’ by NianshengTang and Yuanyuan Ju

Feature screening with missing data is a critical problem but has not been well addressed in theliterature. In this discussion we propose a new screening index based on “information value” andapply it to feature screening with missing covariates.

Asymptotic Theory for Relative-Risk Models with Missing Time-Dependent Covariates

Relative-risk models are often used to characterize the relationship between survival time and time-dependent covariates. When the covariates are observed, the estimation and asymptotic theory for parameters of interest are available; challenges remain when missingness occurs. A popular approach at hand is to jointly model survival data and longitudinal data. This seems efficient, in making use of more information, but the rigorous theoretical studies have long been ignored. For both additive risk models and relative-risk models, we consider the missing data nonignorable. Under general regularity conditions, we prove asymptotic normality for the nonparametric maximum likelihood estimators.

Projection-based Consistent Test for Linear Regression Model with Missing Response and Covariates

In recent years,there has been a large amount of literature on missing data.Most of them focus on situations where there is only missingness in response or covariate.In this paper,we consider the adequacy check for the linear regression model with the response and covariates missing simultaneously.We apply model adjustment and inverse probability weighting methods to deal with the missingness of response and covariate,respectively.In order to avoid the curse of dimension,we propose an empirical process test with the linear indicator weighting function.The asymptotic properties of the proposed test under the null,local and global alternative hypothe tical models are rigorously investigated.A consisten t wild boot strap method is developed to approximate the critical value.Finally,simulation studies and real data analysis are performed to show that the proposed method performed well.

An embedded estimating equation for the additive risk model with biased-sampling data

This paper presents a novel class of semiparametric estimating functions for the additive model with right-censored data that are obtained from general biased-sampling. The new estimator can be obtained using a weighted estimating equation for the covariate coefficients, by embedding the biased-sampling data into left-truncated and right-censored data. The asymptotic properties(consistency and asymptotic normality) of the proposed estimator are derived via the modern empirical processes theory. Based on the cumulative residual processes, we also propose graphical and numerical methods to assess the adequacy of the additive risk model.The good finite-sample performance of the proposed estimator is demonstrated by simulation studies and two applications of real datasets.