Skip to main content
Fig. 1 | BMC Medical Research Methodology

Fig. 1

From: Treatment of missing data in Bayesian network structure learning: an application to linked biomedical and social survey data

Fig. 1

Schematic diagram of Multiple Imputation by Chained Equations approach. For a given incomplete dataset, MICE firstly imputes all missing values via univariate imputation methods. Then it removes the imputed values from variables one by one and creates a model by using the other complete samples. After that, it imputes missingness in each variable in turn using the created model and the remaining variables. These steps are repeated until the data is completed. It then subtracts this new completed data from the initial imputed values to get a difference matrix. The new completed data then becomes the starting point for the next iteration. The whole process is iterated until a pre-defined threshold on the difference between initial imputed and new completed data is met

Back to article page