“…The methods based on empirical likelihood and model‐calibration have inspired further developments on improving the efficiency and robustness against model misspecification. One particular development is the so‐called multiply robust estimators; see Han & Wang (2013), Chan & Yam (2014), Han (2014a, 2014b, 2016a, 2016b), Chen & Haziza (2017, 2018, 2019), Duan & Yin (2017) & Li et al (2020), among others. The model‐calibration approach allows multiple working models for the propensity score and multiple working models for the outcome regression to be accommodated simultaneously in a way similar to the constraints in (20) and (18).…”