mixgb: Multiple Imputation Through XGBoost
Dec 1, 2025·
·
0 min read

Yongshi Deng
Abstract
Multiple imputation is widely used for handling missing data. Traditional methods often require proper model specification to perform well, and can be computationally demanding for large and complex datasets. In this talk, I will introduce our R package mixgb, which provides a scalable and automated solution for multiple imputation by leveraging XGBoost, subsampling, and predictive mean matching. I will discuss how to evaluate multiple imputation methods using simulation studies and I will demonstrate how our visual diagnostic package vismi can be used to assess imputation quality.
Date
Dec 1, 2025 1:30 PM — Dec 5, 2025 3:10 PM
Event
Location
Perth, Australia
Curtin University, Perth, Western Australia