Information Technology Reference
In-Depth Information
Ta b l e 7 . 1 .
Post hoc pairwise comparisons
Abbreviation
Name of test
Distribution
Group size
Strategy
Bonferroni
Bonferroni
t
t
Unequal
Single step
Duncan
Duncan Multiple Range
Studentized range
Equal
Step-down
Gabriel
Studentized
Unequal
Single step
Maximum
Modulus
Hochberg's GT2
Studentized
Unequal
Single step
Maximum
Modulus
LSD
Fisher's Least Significant Difference
t
Unequal
Single step
R-E-G-W-F
Ryan-Einot-Gabriel-Welch
F
F
Equal
Step-down
R-E-G-W-Q
Ryan-Einot-Gabriel-Welch
Studentized Range
Studentized range
Equal
Step-down
Scheffe
F
Unequal
Single step
Sid ak
Sid ak
t
t
Unequal
Single step
S-N-K
Student-Newman-Keuls Multiple
Range
Studentized range
Equal
Step-down
Tukey
Tukey Honestly Significant Difference
Studentized range
Equal
Single step
Tukey-b
Studentized range
Equal
Step down
7.8.3.1 Homogeneity of Variance Assumed
Table 7.1 presents some of the characteristics that distinguish the majority
of tests that assume homogeneity of variance. The tests in Table 7.1 are
listed alphabetically. The columns present the following information:
Abbreviation
:
The abbreviation appearing in the SPSS dialog window
(similar abbreviations are used by SAS).
Name
:
The more complete name of the test.
Distribution
:
The distribution on which the test is based.
Group Size
:
Tests that require equal sample sizes to be performed
by hand as their creators intended are noted as
Equal;
SPSS and SAS will compensate for unequal group sizes
so that the procedures will run under these latter condi-
tions. Tests that were designed to accommodate unequal
group sizes are noted as
Unequal;
these tests are capable
of working with equal sample sizes as well.
Strategy
:
A step-down procedure is only applicable to tests that use
the Studentized range statistic. When we described the
S-N-K procedure in Section 7.4.1.2, we indicated that it
repeated its computation each time as it stepped through
different values of the range. Thus, means separated by
four steps are evaluated differently than those separated
by three steps and so on. The procedure starts with the
largest range (the means that were most different) and
works its way down the line. This is a
step-down
process.
Note that the Tukey HSD test also uses the Studentized
range statistic but does not use a step-down strategy - all
