Missing the Point: Non-Convergence in Iterative Imputation Algorithms

TL;DR

This paper investigates the convergence behavior of iterative imputation algorithms, revealing that valid inferences often occur well before traditional diagnostic indicators suggest convergence, thus questioning the necessity of prolonged iterations.

Contribution

It provides new insights into the convergence properties of iterative imputation, showing that valid inferences are achieved earlier than commonly assumed.

Findings

Inferential validity often occurs after 5-10 iterations.

Traditional diagnostics may overestimate the number of iterations needed.

Longer iterations rarely improve inference quality.

Abstract

Iterative imputation is a popular tool to accommodate missing data. While it is widely accepted that valid inferences can be obtained with this technique, these inferences all rely on algorithmic convergence. There is no consensus on how to evaluate the convergence properties of the method. Our study provides insight into identifying non-convergence in iterative imputation algorithms. We found that--in the cases considered--inferential validity was achieved after five to ten iterations, much earlier than indicated by diagnostic methods. We conclude that it never hurts to iterate longer, but such calculations hardly bring added value.