Marginal analysis of longitudinal ordinal data with misclassification in both response and covariates.

摘要: Marginal methods have been widely used for the analysis of longitudinal ordinal and categorical data. These models do not require full parametric assumptions on joint distribution repeated response measurements but only specify marginal or even association structures. However, inference results obtained from these often incur serious bias when variables are subject to error. In this paper, we tackle problem that misclassification exists in both covariate variables. We develop a method adjustment, which utilizes second-order estimating functions functional modeling approach, can yield consistent estimates valid mean parameters. propose two-stage estimation approach cases validation data available. Our simulation studies show good performance proposed under variety settings. Although is phrased with design, it also applies correlated arising clustered family studies, parameters may be scientific interest. The applied analyze dataset Framingham Heart Study as an illustration.