It is commonly encountered in many applications that variables are not measured perfectly, which is known as errors in variables (EIV) in the statistical literature. It has been long recognized that simply ignoring EIV leads to misleading inference results. There has been extensive work focusing on EIV in explanatory variables only. EIV in both response and explanatory variables has received very limited attention especially when they are discrete/categorical. Our work focuses on the joint misclassification errors in a binary response variable and a binary explanatory variable. To account for the possible dependence between the misclassification errors in both variables, we introduce the dependence parameters following the notion in Vogel et al. (2005). We conduct sensitivity analysis to check the consequence of fitting an independent misclassification model to data actually generated from dependent misclassification models.