Missing data imputation in PLS-SEM

As a useful tool for business research, PLS-SEM is widely adopted for the assessment of causal-predictive relationships of models when developing and testing theories. Nevertheless, the less error-prone techniques for handling missing data are routinely ignored by PLS-SEM researchers. In this paper, we propose an imputation method, called EM PLS-SEM, to deal with missing values in PLS-SEM. The method takes advantages of the estimation procedure of PLS-SEM to reach the goal of filling the missing elements with values that are most likely to appear. Numerical studies verify that the proposed method outperforms other alternatives in data completion and model fitting.