You use the BLOCKS statement to find a D-optimal design in the presence of fixed covariates (for example, blocks) or covariance.
The technique is an extension of the optimal blocking technique of Cook and Nachtsheim (1989); see the section Optimal Blocking.
For the purposes of optimal blocking, the model for the original candidate points is referred to as the treatment model; the candidate points for the part of the design matrix corresponding to the treatment model form the treatment set.
If the GENERATE statement is not specified, then the full candidate set is used as the treatment set; otherwise, an optimal
design for the treatment model ignoring the blocks is first generated, and the result is used as the treatment set for optimal
blocking.
The following are three mutually exclusive block-specifications that you can provide:
-
COVAR=SAS-data-set VAR=( variables )
-
specifies a data set to use in providing a general covariance matrix for the runs. The argument to VAR= names the variables
in this data set that contain the columns of the covariance matrix for the runs. For an example, see Example 14.9.
-
DESIGN=SAS-data-set
-
specifies a data set to use in providing a general covariate model. In addition to this data set, you must specify a covariate
model with the CLASS and MODEL statements. Covariate models are specified in the same way as the treatment model; CLASS and
MODEL statements that come after a BLOCKS statement involving the DESIGN= specification are interpreted as applying to the
covariate model. For an example, see Example 14.8.
-
STRUCTURE=(b) k
-
specifies a block design with b blocks of size k. For an example, see Example 14.7.
The following options can also be used:
-
INIT=RANDOM
-
specifies the initialization method for constructing the starting design. The option INIT=RANDOM specifies that the starting
design is to be constructed by selecting candidates at random without replacement. The option INIT=CHAIN selects candidate
points in the order in which they occur in the original data set.
-
ITER=n
-
specifies the number of times to repeat the search from different initial designs. Because local optima are common in difficult
search problems, it is often a good idea to make several tries for the optimal design with a random or partially random method
of initialization (see the preceding INIT= option). By default, n = 10. Specify both INIT=CHAIN and ITER=0 to evaluate the initial design itself.
-
KEEP=m
-
specifies that only the best m designs are to be retained. The value m must be less than or equal to the value n of the ITER= option. By default m = n, so that all iterations are kept. This option is useful when you want to make many searches to overcome the problem of local
optima but you are only interested in the results of the best m designs.
-
NOEXCHANGE
-
suppresses the part of the optimal blocking algorithm that exchanges treatment design points for candidate treatment points.
When this option is specified, only interchanges between design points are performed. Use this option when you do not want
to change which treatment points are included in the design and you only want to find their optimal ordering.