This principle states that the missing information in data with missing values is equal to the difference between the information matrix for complete data and the observed information matrix.