Imputation Based on Propensity Score: A Comparative Study of the Penalized Spline Propensity Prediction Method with Alternative Double Robust Estimators

• Activity Number: 468
• Type: Contributed
• Date/Time: Wednesday, August 1, 2007 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Guangyu Zhang*+ and Roderick J. Little
• Companies: University of Michigan and University of Michigan
• Keywords: missing data ; imputation ; PSPP ; propensity score

Abstract

Little and An (2004) proposed the penalized spline propensity prediction method of imputation (PSPP). The predicted unconditional mean of the missing variable has a double robustness (DR) property under mis-specification of the imputation model. The DR property can also be achieved by modeling the relationship parametrically. One method is to include the inverse of the propensity score as a linear term in the imputation model (Firth and Bennett, 1998; Bang and Robins, 2005). Another approach is to calibrate the predictions from a parametric model by adding means of the weighted residuals, with weights equal to inverse of the propensity scores (Robins, Rotnitzky and Zhao, 1994; Scharfstein, Rotnitzky and Robins, 1999). In this talk, I compare the PSPP method with these methods by simulation.