Linear regression through the origin with constant coefficient of variation for the inverse gaussian distribution

Abstract

A simple linear regression model with no intercept term for the situation where the response variable obeys an inverse Gaussian distribution and the coefficient of variation is an unknown constant is discussed. Maximum likelihood estimators and the confidence limits of the regression parameter are obtained. Finally uniformly minimum variance unbiased estimators of parameters are given.

Keywords:

• zero intercept
• maximum likelihood estimator
• uniformly minimum variance unbiased estimator
• confidence limits