The estimated treatment effects (group differences) that were common to all
follow-up times are shown in Table 2. Analyses of each of the specified outcomes
during the treatment phase (6, 12 and 24 months) only, using the repeated
measures (random effects) model, produced significant group differences for
only one of the outcomes: YSQ defectiveness/shame schema (Table 2), where
the score increased. Repeating the analysis after the inclusion of the 36-month
follow-up data did not change the interpretation of the results. In view of the
preliminary nature of the analysis and the many tests being carried out, the statistically
significant result may be simply due to chance. Applying a Bonferroni
correction to control for familywise error showed that a significance level where
p < .003 would be required for us to be confident that this was not a type 1 error.
As p = .008 in this analysis, then a type 1 false positive error cannot be ruled out.
There were clinically significant improvements reported in the right direction
on all the YSQ schemata between baseline and 36-month follow-up, except for the ‘abandonment’, ‘self-sacrificing’ and ‘defectiveness/shame’ schema. However,
none of these changes were statistically significant.