Influence of parallel computing strategies of iterative imputation of missing data: a case study on missForest

TL;DR

This paper investigates how two parallel computing strategies affect the accuracy and statistical properties of the missForest imputation method for missing data, revealing differences in bias and results.

Contribution

It provides a comparative analysis of variable-wise and model-wise parallel strategies in missForest, highlighting their impacts on imputation bias and statistical estimates.

Findings

Both strategies yield similar prediction errors.

Variable-wise strategy introduces additional biases.

Parallel strategies influence statistical estimates differently.

Abstract

Machine learning iterative imputation methods have been well accepted by researchers for imputing missing data, but they can be time-consuming when handling large datasets. To overcome this drawback, parallel computing strategies have been proposed but their impact on imputation results and subsequent statistical analyses are relatively unknown. This study examines the two parallel strategies (variable-wise distributed computation and model-wise distributed computation) implemented in the random-forest imputation method, missForest. Results from the simulation experiments showed that the two parallel strategies can influence both the imputation process and the final imputation results differently. Specifically, even though both strategies produced similar normalized root mean squared prediction errors, the variable-wise distributed strategy led to additional biases when estimating the mean and inter-correlation of the covariates and their regression coefficients.