ABSTRACT
Regression coefficients are like the expectation of a random variable location parameter. Therefore, we can use the methodology from Chapter two to determine the sample sizes needed for a given precision. But there are two totally different situations concerning the underlying model. This fact is seldom recognised in analysing data because the standard software packages, contrary to the situation in ANOVA, make no difference between the models when determining regression analysis. The reason is that the least squares approach leads to the same formula for both models for estimating the regression coefficients, but the standard errors are different for these estimators in the two models.
