All experiments at one level

The sum of squares of residuals, with three degrees of freedom, is given in this case by 

The sum of squares due to lack of fit, with one degree of freedom, may be obtained by difference 

An appropriate test for lack of fit is the / -test described in Section 6.5. 

The critical value of F 2) at the 95% level of confidence is 18.51 (see Appendix C). Thus, F which allows us to state (with a risk of being wrong no more than one time in 20) that the model = 0 + r,. exhibits a significant lack of fit to the three data we have obtained. The level at which F = F is 97.8% confidence, or a risk a = 0.022. A better model is probably possible. 

If we fit the model y,. = Po + r, to this data, we will find that = 3.9. The sum of squares of residuals (with two degrees of freedom) is 

---

Although it is not recommended, frequently, for practical reasons, experiments are blocked or sorted by one or more factors (4, 5,16,106). Then all experiments at one level of the factor are executed hrst, followed by all at the other level. The experiments are usually performed in a random sequence within one block. At least before and after each block, a check for drift can be recommended in such situation. These latter experiments permit observation and correction of occasional block effects. 

---