2.1. LACK OF FIT AND PURE ERROR 49

Apuntes

2.1. LACK OF FIT AND PURE ERROR 49 which is the (T2 of our problem. In chemical experiments, a succession of readings made during steady-state running does not provide genuine repeat points. However, if a certain set of conditions was reset anew, after intermediate runs at other X-levels had been made, and provided drifts in the response level had not occurred, genuine repeat runs would be obtained. With this in mind, repeat runs that show remarkable agreement which is contrary to expectation should always be regarded cautiously and subjected to additional investigation. Calculation of Pure Error and Lack of Fit Mean Squares When there are repeat runs in the data, we need additional notation to take care of the multiple observations on Y at the same value of X. Suppose we have m different values of X and, at the jth of these m particular values, Xj' where j = 1, 2, ... , m, there are nj observations; we say that: Y jj , Y 12 , .•• , Yjn, are nj repeat observations at Xl; Y 21 , Y 22 , •.• , Y 2 1!2 are n2 repeat observations at X 2 ; Y ju is th...