Approximate subject-deletion influence diagnostics for Inverse Probability of Censoring Weighted (IPCW) method

An approximate formula for subject-deletion influence diagnostics is proposed for the Inverse Probability of Censoring Weighted method [Robins, J.M., Rotnitzky, A., Zhao, P., 1995. Analysis of semiparametric regression models for repeated outcomes in the presence of missing data. J. Amer. Statist. Assoc. 90, pp. 106–121] when the independent working correlation is employed. By a numerical study with a dataset from a clinical trial, it is found that the formula provides good approximation to the exact method by fitting regression models repeatedly to datasets without each subject and saves the computational time remarkably in particular for large datasets.