A new method has been developed to identify and estimate mismeasured and endogenous regressor models in situations where other methods are not available. This method uses a heteroskedastic covariance restriction that is common in models involving mismeasurement or endogeneity. The researchers have shown that this method can be applied to various types of models, including partly linear models. They have also provided estimators for these models. In cases where specific assumptions do not hold, set identification bounds have been derived. An example application of this method to estimate Engel curves has been presented.