Infer — Entropy-Based Imputations

Under Infer, we have two additional options, namely Entropy-Based Static Imputation and Entropy-Based Dynamic Imputation. As their names imply, they are based on Static Imputation and Dynamic Imputation.

Whereas the standard (non-entropy-based) approaches randomly choose the sequence in which missing values are imputed within a row of data, the entropy-based methods select the order based on the conditional uncertainty associated with the unobserved variable. More specifically, missing values are imputed first for those variables that meet the following conditions:

1. Variables that have a fully-observed/imputed Markov Blanket;

2. Variables that have the lowest conditional entropy, given the observations and imputed values.

The advantages of the entropy-based methods are (a) the speed improvement over their corresponding standard methods and (b) their improved ability to handle datasets with large proportions of missing values.