seva SPLIT-PLOT DESIGN o —— nn Is appropri: e for many research situations. For example, in the vigilance experiment, subjects assigned to the auditory display condition a, might find this m« de of signal presentation unpleasant. As a result, two of the subjects mig :t refuse to complete the experiment. In this example, unegual cell freguen. xs result from the nature of the experimental treatments. A least-sguare . analysis, rather than an unweighted-means analysis, should be used. Computational formulas based on the data in Table 8.10-1 are given in Table 8.10-3. TABLE 8.10-3 (Computational Procedures for Least-Sguares Solution for Type SPF-2.4 Design SS,naj < [ABS] — [X] < 235.500 Stkerwcen subj — [4S] - [X] < 12.500 SS, — [4] — [X] < 5.633 SS.uni wgroups — LAS] — [A] < 6.867 SS, iunin swej < [ABS] — [AS] < 223.000 SS, — [B] — [X] — 194.500 SS,e — [AB] — [A] — [B] £ [X] — 15.634 SS. vni warovps — [ABS] — [AB] — [AS] t [A] < 12.866 The analy: s is summarized in Table 8.10-4. In a least-sguares analysis, the partitioned sum of sguares add up to the total sum of sguares. Tests of simple main effects and comparisons among means have the same general form as tests based on egual cell freguencies. These procedures are illustrated in Sections 8.6 and 8.7. They generalize to the least-sguares solution but reguire the substitution of the appropriate value for n. TABLE 8.10-4 Analysis of Variance Table for Least- Sguares Solution Source SS df MS F |. Between subjects 12.500 N-1:7 2 A 5.633 p-i