1: \begin{abstract}

2:   We present a second-order estimator of the mean of a variable subject to

3:   missingness, under the missing at random assumption. The estimator improves

4:   upon existing methods by using an approximate second-order expansion of the

5:   parameter functional, in addition to the first-order expansion employed by

6:   standard doubly robust methods. This results in weaker assumptions about the

7:   convergence rates necessary to establish consistency, local efficiency, and

8:   asymptotic linearity. The general estimation strategy is developed under the

9:   targeted minimum loss-based estimation (TMLE) framework. We present a

10:   simulation comparing the sensitivity of the first and second order estimators

11:   to the convergence rate of the initial estimators of the outcome regression

12:   and missingness score. In our simulation, the second-order TMLE improved the

13:   coverage probability of a confidence interval by up to $85\%$. In addition, we

14:   present a first-order estimator inspired by a second-order expansion of the

15:   parameter functional. This estimator only requires one-dimensional smoothing,

16:   whereas implementation of the second-order TMLE generally requires kernel

17:   smoothing on the covariate space. The first-order estimator proposed is

18:   expected to have improved finite sample performance compared to existing

19:   first-order estimators. In our simulations, the proposed first-order estimator

20:   improved the coverage probability by up to $90\%$. We provide an illustration

21:   of our methods using a publicly available dataset to determine the effect of

22:   an anticoagulant on health outcomes of patients undergoing percutaneous

23:   coronary intervention. We provide R code implementing the proposed estimator.

24: \end{abstract}

25: