For Multiple x, Variables

The same basic formula is used (Equation 8.6). The — y = e, values are taken from the entire or full model, but the ej values are regressed only on the Xj predictor variables to be evaluated or, if the entire model is used, all are regressed. 

Using the data from Table 8.2 and all x, values, the regression equation is  

the researcher probably would be very interested in the increasing variance in this example. The data suggest that, as time goes by, not only does the potency diminish, but also with increasing variability. This could flag the researcher to sense a very serious stability problem. In this case, transforming the data to stabilize the variance may not be useful. That is, there should be a 

Before proceeding to the weighted least squares method, we need to discuss a basic statistical procedure that will be used in weighted regression. 

---