Data analysis
The data analysis was conducted on an intention-to-treat basis that maintained the advantages of random allocation that might otherwise have been lost if participants were excluded from the analysis because they withdrew from the study or failed to comply with treatment (26). Pre- and posttest scores and demographic data were analyzed by using SPSS for Windows version 11.0. Demographic differences between the three groups were assessed with analysis of variance (ANOVA) or the Kruskal-Wallis test by ranks (H statistic), as appropriate. Multivariate analyses of variance (MANOVAs) (3 × 3) were performed for the dependent variables to determine whether treatment was associated with the interactive effects postulated (group × time). Post-hoc analysis using Tukey's honestly significant difference test for multiple comparisons was performed for measures for which results were significant in the MANOVA (27). The level of significance for the statistical tests was set at .05.