A dynamic Cholesky data imputation method for correlation structure consistency

In the context of data that is missing completely at random, we propose a new data imputation method that exploits Cholesky decomposition. The data imputation method falls within the multiple imputation framework and is designed to ensure consistency with the correlation structure of the available data. The advantage is an accessible and computationally efficient approach to managing missing data that avoids the model risk associated with applying complex model-based data imputation methods. The non-recursive nature of our data imputation method further avoids the convergence issues associated with recursive approaches.