DefinePK

Abstract

Incomplete data is a frequent occurrence in many research areas especially cross sectional survey data in epidemiology, health and social sciences research. In this paper, the effect of missing observations were accounted for by using multiple imputation (MI) and inverse probability weighting (IPW) methods. Generally, multiple imputation has the ability to draw multiple values from plausible predictive distribution for the missing values. However, under the inverse probability weighting procedure the weights are the inverse of the predicted probabilities of response estimated from the missing ness models of incomplete variables. A simulation study is conducted to compare methods and demonstrate that a cross sectional survey data can be used to mitigate bias induced by missing data. The application and simulation results show the benefit of the IPW compared with the MI. The former performs well but not as the latter.