Prediction of Gross Tree Volume Using Regression Models with Non-Normal Error Distributions

摘要

Previous work in weighted linear regression, where weight functions are used to obtain homogeneous variance on a transformed scale, has often assumed that the errors are normally distributed. In a study of four data sets, three of which were actual data sets with unknown error distributions and one an artificial set with a known error distribution, this assumption is incorrect. Consequently, we tested a transformation of the normal distribution, called the S_U distribution, and compared it with the normal as an alternative. Forthree of the four data sets studied, the S_U distribution was superior. Prediction intervals and biases forthe regression estimators generated usingthe S_U and normal distributions were also evaluated. Results for the S_U distribution bettered those for the normal distribution in three of the four data sets. Forthe remaining data set, they were comparable.