Errors in the factor levels and estimation of regression model parameters

Abstract

The regression model nonlinear in factors and having errors in the factor levels settings, is transformed into a model with all the errors being in the response. A method for estimation of the transformed model parameters is proposed provided that the moments of the errors in the factor levels are known. The properties of the obtained estimates are studied. The proposed estimates are compared with other methods on the basis of an example.

Keywords:

• Regression
• weighted least squares
• heteroscedasticity