In longitudinal data analysis with dropouts, despite its local efficiency in theory, the augmented inverse probability weighted (AIPW) estimator hardly achieves the semiparametric efficiency bound in practice, even if the variance–covariance of the longitudinal outcomes is correctly modeled....

The augmented inverse probability weighted (AIPW) estimator employing the optimal augmentation term is more efficient than the inverse probability weighted (IPW) estimator. However, the AIPW estimator could lose substantial efficiency compared to the IPW estimator when the optimal augmentation...