Abstract
Measurement error and misclassification in covariates is always inevitable in data collection process. Treating variables measured with error as true value to build up regression models and perform statistical analysis can lead to misleading results. To get more accurate estimation results, the effect of measurement error and misclassification has to be considered. In this thesis, we mention some correction methods to estimate the parameters in linear and logistic regressions on error-prone covariates and one two-level categorical variable with misclassification. Simulation studies are made to compare the performance of these estimators for a finite sample. Finally, we present a real-data example on logistic regression.