Marianne Welmers

Chapter 5 118 problem scale items, as well as items referring to other problems such as thought and attention problems. Because the number of items per scale (32 for internalizing and 35 for externalizing problems) outrated the number of participating families, we reasoned that a proper estimate of reliability was not viable and thus we did not calculate Cronbach’s alpha for the current study sample. The manual of the Dutch CBCL version reports good reliability and validity (Van der Hulst, & Van der Ende, 2013). Statistical Analyses To investigate unbalanced alliances within the family, we calculated difference scores for alliances of two family members. When there were three family members with alliance scoreswithin thesame family ( n =7 for therapist reports, n =2 for self- andobserver reports) we used scores of the two family members that were most discrepant. To investigate the effect of unbalanced alliances within the family, and the family’s shared sense of purpose on youth internalizing, externalizing, and total behavior problems at follow-up, we first calculated zero order correlations (pearson’s r ) between the included alliance measures (alliance difference scores and SSP) and follow-up CBCL scores. Because of the small sample size, we also performed post hoc non-parametric tests (Spearman’s rho ) for all significant effects we found. Next, to control for youth internalizing, externalizing, and total problems at the start of treatment, we calculated partial correlations for all alliance measures and follow up CBCL scores, controlling for CBCL scores at the start of treatment. Because of the explorative nature of the study and the small sample size with limited statistical power, we also reported trend-significant and substantial non-significant correlations that might be of theoretical importance. Results Preliminary Analyses First, we performed a paired t-test with baseline and follow-up CBCL scores (Table 1), to test whether youth problems improved from intake to 18 months post-treatment. We found that externalizing youth problems ( t = 5.322, p < .001) and total youth problems ( t = 4.212, p < .001) improved significantly from the start of treatment to 18 months after ending treatment. For internalizing problems, the t-test was not significant, but indicated a trend ( t = 1.797, p = .083).