The routines described in this chapter are for commonly-used experimental designs. Typically, responses are stored in the input vector Y in a pattern that takes advantage of the balanced design structure. Consequently, the full set of model subscripts is not needed to identify each response. The routines assume the usual pattern, which requires that the last model subscript change most rapidly, the next to last model subscript change next most rapidly, and so forth, with the first subscript changing the slowest. This pattern is referred to as lexicographical ordering.

Routines AONEW, AONEC, and ANEST allow missing responses. NaN (not a number) is the missing value code used by these routines. Use routine AMACH to retrieve NaN. Any element of Y that is missing must be set to AMACH(6). For a description of AMACH, see the section Machine-Dependent Constants in the Reference Material Other routines described in this chapter do not allow missing responses because they generally deal with balanced designs.

As a diagnostic tool for determination of the validity of a model, routines in this chapter typically perform a test for lack of fit when n (n > 1) responses are available in each cell of the experimental design. Routines in Chapter 2, “Regression,” are useful for analysis of generalizations of many of the models treated in this chapter. In particular, Chapter 2 provides routines for the general linear model.