Parameters involved in a linear model are estimated when both independent and dependent variables in a replicated case are subject to error. It is also assumed that errors are normally distributed with mean zero and two cases with common and different variances are discussed.