The difference between the predicted and observed values of a model due to random error including missing and incomplete data.