ABSTRACT
This chapter introduces a class of functions, namely the estimable functions, for which the least squares estimator is unbiased. It discusses how these assumptions can help us determine desirable estimation properties for linear models. While unbiasedness is generally a desirable property for an estimator, it does not guarantee that it is the "best" estimator. The program used, called GLM, is a general linear model program and uses a generalized inverse to solve the normal equations. In order to proceed further in statistical inference for linear models, one need to become acquainted with the multivariate normal distribution and several basic theorems associated with this distribution. The assumption of normality will facilitate the making of tests and the construction of confidence intervals for linear model parameters. The linear model having these assumptions is sometimes referred to as the general or classical linear model.
