Itemwise conditionally independent nonresponse modeling for incomplete multivariate data

TL;DR

This paper proposes a new nonresponse mechanism for multivariate data where each variable's missingness depends on other variables, allowing for flexible modeling and sensitivity analysis of nonignorable missing data.

Contribution

It introduces an itemwise conditionally independent nonresponse model that is identifiable and applicable to nonignorable missing data, with methods for sensitivity analysis.

Findings

Model is identifiable under the proposed assumptions.

Sensitivity analysis methods are developed.

Illustrative data analyses demonstrate practical use.

Abstract

We introduce a nonresponse mechanism for multivariate missing data in which each study variable and its nonresponse indicator are conditionally independent given the remaining variables and their nonresponse indicators. This is a nonignorable missingness mechanism, in that nonresponse for any item can depend on values of other items that are themselves missing. We show that, under this itemwise conditionally independent nonresponse assumption, one can define and identify nonparametric saturated classes of joint multivariate models for the study variables and their missingness indicators. We also show how to perform sensitivity analysis to violations of the conditional independence assumptions encoded by this missingness mechanism. Throughout, we illustrate the use of this modeling approach with data analyses.