Deep Generative Imputation Model for Missing Not At Random Data

By J. Chen et al

Published on June 10, 2023

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Problem Setting
3. Non-negligible Missing Mechanism Modeling
3.1 Existing MNAR modeling approaches
3.1.1 Serial-structure models
3.2 Parallel-structure conjunction model

Summary

The document discusses a deep generative imputation model for handling Missing Not At Random (MNAR) data. It emphasizes the importance of modeling the joint distribution of complete data and missing mask to improve imputation accuracy. Existing methods like pattern mixture model, selection model, and pattern-set mixture model are critiqued for their limitations. The proposed conjunction model introduces an auxiliary variable for joint representation of multimodal missing data. This new approach aims to avoid information bottlenecks and biases in imputation models.