Estimation and Testing with Normal Errors
From Jane Harvill
In this lesson, the assumption of the normality of errors, and hence the normality of the responses is added to the simple linear regression model. The consequences are that maximum likelihood estimators can be derived. Additionally, sampling distributions of parameter estimates can be obtained and used for inference (testing and confidence intervals) on the population parameters. This also leads to a method in regression that is analogous to the ANOVA test for equal treatment means.