Sensitivity Analysis for Non-ignorable Missing Data With Misclassification Error
Presentation given by Georgia Southern Faculty members Rochani D. Haresh, Hani M. Samawi, and Yu Lili at the 2018 American Public Health Association (APHA) Meeting.
Missing data and misclassification error are very common problem in many research studies. It is well known that the misclassification error in covariates can cause bias estimation of parameters for statistical model as well as it can reduce the overall statistical power. Misclassification simulation extrapolation (MC-SIMEX) procedure is a well-known method to correct the bias in parameter estimation due to misclassification for given statistical model. The misclassification matrix has to be known or estimated from a validation study to use MC-SIMEX method. However, in many circumstances, the validation study has non-ignorable missing data. Estimation of misclassification matrix can be biased due to presence of non-ignorable missing data which lead to bias estimation of parameters for our statistical model. In this paper, we apply the Baker, Rosenberger and Dersimonian modeling approach to perform the sensitivity analysis using MC-SIMEX method. Simulation studies are used to investigate the efficiency of parameters under given assumption of missing data mechanism. We illustrate the method by application to the “National Health and Nutrition Examination Survey” dataset.
American Public Health Association (APHA) Meeting
San Diego, CA
Rochani, Haresh, Hani Samawi, Lili Yu.
"Sensitivity Analysis for Non-ignorable Missing Data With Misclassification Error."
Biostatistics, Epidemiology, and Environmental Health Sciences Faculty Presentations.