dlglm

dlglm fits a regression model in the presence of missing data. Specifically, it is able to handle any mechanism of missing data, including the difficult Missing Not At Random (MNAR) case. By utilizing neural networks, dlglm can learn complex dependencies across covariates and the response of potentially high-dimensional datasets. Additionally, dlglm uses a neural network to model the missingness mechanism, which can be adapted to account for any Missing Completely at Random (MCAR), Missing at Random (MAR), or MNAR missingness. By increasing the complexity of the missingness neural network, potentially complex dependencies between the missingness of the features can also be captured. Additionally, dlglm is built in the supervised learning setting, allowing for prediction on a trained model.