Statistical Methods for Psychology

(Michael S) #1
12.9 Computer Solutions 401

Means with the same letter are not significantly different.
REGWQ Grouping Mean N CONDTION
A 29.000 8 Mc-M
A 24.000 8 S-M
B 11.000 8 S-S
B 10.000 8 M-M
B 4.000 8 M-S


Tukey’sStudentized Range (HSD) for variable: LATENCY
NOTE: This test controls the type I experimentwise error rate, but generally has a higher
Type II error rate than REGWQ.
Alpha = 0.05 df = 35 MSE = 32
Critical Value of Studentized Range = 4.066
Minimum Significant Difference = 8.1319 Critical range for all differences
Means with the same letter are not significantly different.
Tukey Grouping Mean N CONDTION
A 29.000 8 Mc-M
A 24.000 8 S-M
B 11.000 8 S-S
B 10.000 8 M-M
B 4.000 8 M-S


Scheffe’stest for variable: LATENCY
NOTE: This test controls the Type I experimentwise error rate but generally has a higher
Type II error rate than REGWQ for all pairwise comparisons
Alpha = 0.05 df = 35 MSE = 32
Critical Value of F = 2.64147
Minimum Significant Difference = 9.1939 Critical range for all differences
Means with the same letter are not significantly different.
Scheffe Grouping Mean N CONDTION
A 29.000 8 Mc-M
A 24.000 8 S-M
B 11.000 8 S-S
B 10.000 8 M-M
B 4.000 8 M-S


Exhibit 12.1 (continued)


regardless of where they lie in an ordered series. For this specific set of data our conclu-
sions are the same as they were for the Ryan test, although that will certainly not always be
the case.
Although the Scheffé test is run quite differently from the others, it is possible to com-
pute a critical range for all pairwise comparisons. From Exhibit 12.1 we can see that this
range is 9.1939, almost a full point larger than the critical range for Tukey. This reflects the
extreme conservatism of the Scheffé procedure, especially with just pairwise contrasts, and
illustrates my major objection to the use of this test for this purpose.
Free download pdf