A Step by Step Guide for SPSS and Exercise Studies

78 Statistical tests
have stopped here and reported that the three styles did not differ significantly in
clearance height.
The post-hoc Tukey test compares the clearance height of the three styles.
Note that the use of the Tukey test is justified because the Levene Test was not
significant (the actual value of the test is not shown here). If the Levene Test
was significant, you should have used one of the post-hoc tests under the equal
variances not assumed section of Dialog box 67. Table 24 shows that the
clearance height differs significantly among all three styles. For example,
when using the Fosbury style the clearance height is 13.76 cm and 5.36 cm
smaller than the clearance height obtained with the Western Roll and the
Straddle styles respectively. Table 24 shows the mean difference in clearance
height, as well its standard error, significance level and 95% confidence
intervals. These intervals show the values two standard errors below (Lower
Bound) and above (Upper Bound) the mean difference respectively.
Note that some statisticians (e.g., Pedhazur and Schmelkin, 1991) do not
recommend the use of post-hoc tests, because these tests require a large
number of mean comparisons which can increase the probability for Type I
error (especially if the significance level in Dialog box 67 is not adjusted).
Pedhazur and Schmelkin (1991) advocate the use of a priori planned
comparisons to prevent Type I errors. A priori comparisons perform only a
limited number of comparisons between mean scores, because they are based
on a theory that specifies which comparisons are important and which are not.
For example, the post-hoc Tukey test above carried out three mean
comparisons contrasting each style with the others. If there were five styles,
you would have performed 10 different mean comparisons. However, with a
priori planned comparisons you could limit the comparisons to a certain
number specified by a theory or previous research (e.g., compare the Straddle
and Fosbury styles only).
To carry out a priori planned comparisons, click Contrasts in Dialog box 66
to open Dialog box 68. Select Polynomial and Linear under the Degree option.
Each style should be given a comparison coefficient. The order of the
coefficients is crucial because each coefficient corresponds to a different high
jump style. Although there are many types of planned comparisons, orthogonal
ones are most often used. Orthogonal planned comparisons require that the sum
of the coefficients is zero in any given comparison. That is, if you want to
compare the first and the third style, you should assign coefficient 1 to the first
style (Western Roll), coefficient 0 to the second style (Straddle), and coefficient
1 to the third style (Fosbury) (i.e., 1 ‡ 0 1 ˆ 0). If the first style was compared
with the other two styles, the coefficients for the three styles should have been
2, 1, and 1 respectively. Use the Add button to add each coefficient. For more
complicated designs use the Next button to add another set of contrasts. Click
Continue and when you get back to Dialog box 66 click OK. The output tables
(Tables 25, 26) list the contrast coefficients and the results of the planned
comparisons between the first and the third style. As you can see, the t test is
significant which indicates that the difference in clearance height between the