Searching for and Analyzing Designs with Two-Level Factors - Search Criterion Tab

Select the Search criterion tab of the Searching for and Analyzing Designs with Two-Level Factors dialog to access the options described here.

Criterion for best design

Use the options in the Criterion for best design group box to specify the criterion to use in creating your design.

Maximum resolution (up to R specified below)

If the Maximum resolution (Up to R specified below) option button is selected, design resolution is the sole search criterion. The design search is satisfied when a design is found of maximum resolution or if a design is found of the resolution specified in the Highest Resolution (to search for) box (see below).

Maximum R plus maximum unconfounding

If the Maximum R plus maximum unconfounding option button is selected, Statistica searches for the design of maximum resolution in which the maximum number of interactions of the crucial order, given the resolution, are unconfounded with all other interactions of the crucial order. If no interactions of the crucial order, given the resolution, are unconfounded with all other interactions of the crucial order, Statistica automatically enables the Maximum R plus minimum aberration option (see below).

Maximum R plus minimum aberration

If the Maximum R plus minimum aberration option button is selected, Statistica searches for the design of maximum resolution with minimum aberration. In general, minimum aberration designs have the smallest possible number of pairs of confounded interactions of the crucial order, given the resolution.

Highest Resolution (to search for)

The value in the Highest Resolution (to search for) box specifies the resolution of designs to consider when searching for a design when the Maximum resolution (up to R specified below) search criterion is selected. Enter the resolution of the highest resolution designs you want to include in the search. The design search will be satisfied when a design is found of maximum resolution or if a design is found of the resolution specified in the Highest Resolution (to search for) box. The first such design that is found will be reported, i.e., if the Comprehensive search of all generators option button is selected on the Search Strategy tab, it is disabled when the Maximum resolution (up to R specified below) search criterion is selected.

If either the Maximum R plus maximum unconfounding option button or the Maximum R plus minimum aberration option button is selected, Statistica automatically searches for designs of maximum resolution.

Keep selected effects unconfounded

In addition to, or instead of, the Maximum R plus maximum unconfounding criterion or the Maximum R plus minimum aberration criterion, you can specify particular effects of the crucial order, given the resolution, to keep unconfounded. Select the Keep selected effects unconfounded check box or click the  to display the Unconfounded Effects dialog, in which you select the effects of the crucial order, given the resolution, that you want to keep unconfounded. This option operates by eliminating from the search those designs with effects that are confounded with the specified effects. Because of the nature of the sets of generators that are searched (i.e., non-isomorphic sets, excluding sets where factor names are merely relabeled), the Comprehensive search of all generators option on the Search strategy tab cannot be selected when the Keep selected effects unconfounded option is used.

It is recommended that you conduct a search using the Maximum R plus maximum unconfounding criterion before searching for a design with user-specified unconfounded effects. Inspection of the maximally unconfounded design can show you how many unconfounded effects are possible and whether there are any constraints on the unconfounded effects. For example, in many maximally unconfounded designs the unconfounded effects are constrained to those interactions that involve just one or two factors, say, all the interactions involving factor 1. It would be futile, therefore, to search for a design in which all the interactions involving factor 1 and all the interactions involving factor 2 are unconfounded. If a design cannot be found with the user-specified effects kept unconfounded, an alternative design meeting the other specified criteria is sought.

For more detailed information on search criteria and search options, see 2(k-p) Maximally Unconfounded and Minimum Aberration Designs.