Fig. 2From: Treatment of missing data in Bayesian network structure learning: an application to linked biomedical and social survey dataSchematic diagram of Structural Expectation-Maximization algorithm. SEM has two components: E-step and M-step. It considers a BN structure for the incomplete data at the very beginning. Then it applies the iterative two steps, alternating E-step and M-step. E-step estimates the values of missing data by computing the expected statistics using the current network structure. The M-step maximizes the scoring function and updates the resulting network structure. These two steps are repeated until convergence is metBack to article page