Structural Multilevel Modelling in Combination with Misclassification andMeasurement Error in Covariates

Many multilevel epidemiological frameworks are impressed by measurement error or misclassification in covariates. It has been well established that errors in covariates degrade the quality of statistical inference, lead to biased estimates and a loss of power to detect associations between covariates and the outcome variable. In this paper, sponsoring various features of discrete and continuous error-prone variables, we consider multilevel settings with misclassified and mismeasured covariates. Contemplating structural likelihood-based strategy, we develop an estimation and inference method that accomodates both sources of errors simultaneously using the multivariate Gauss-Hermite quadrature technique to approximate the likelihood function numerically. Simulation results show that the proposed method performs well when correcting for covariate measurement error and non-differential misclassification in terms of bias, empirical standard error, root of mean squared error as well as the coverage rate.