Missing Data Handling via EM and Multiple Imputation in Network Analysis using glasso and atan Regularization

The existing literature on missing data handling in psychological network analysis using cross-sectional data is currently limited to likelihood based approaches. In addition, there is a focus on convex regularization (the graphical lasso), with the missing handling implemented using different calculations in model selection across various packages. To address these shortcomings in the current literature, we (1) implement a direct EM approach to handle missing values in network analysis using nonconvex regularization (specifically using the atan penalty) and a stacked multiple imputation approach suited for atan as well as graphical lasso, (2) standardize model estimation and selection to (3) facilitate evaluating the specific performance of missing data handling techniques. Simulated conditions vary network size, number of observations, and amount of missingness. Evaluation criteria encompass the recovery of the edge set, partial correlation bias, and correlation of network statistics. Overall, missing data handling approaches exhibit similar performance under many conditions. Using graphical lasso with EBIC model selection, the two-step EM method performs best, closely followed by stacked multiple imputation. For atan regularization using BIC model selection, stacked multiple imputation proves most consistent across all conditions and evaluation criteria.