TPC Journal-Vol 10- Issue 3-FULL ISSUE

The Professional Counselor | Volume 10, Issue 3 355 (e.g., there was no statistical difference between the two reports, all values of p > .05). As shown in Table 1, adolescent and caregiver reports only statistically differed for behavioral control ( t [28] = 4.23, p < .001) and communication ( t [28] = 2.96, p = .006). Specifically, adolescents reported higher levels of both behavioral control and communication; these high levels are considered indicative of unhealthy or distressed families (I. W. Miller et al., 1985). Table 1 Descriptive Statistics and Group Comparisons of Family Functioning Variables as Reported by Adolescents and Caregivers Variable Adolescent Caregiver Cut-Off M SD Range M SD Range t (28) p Problem-solving 2.2 2.58 0.65 1.00–3.80 2.29 0.57 1.40–3.80 1.88 .071 Communication 2.2 2.56 0.37 1.83–3.17 2.30 0.37 1.50–3.00 2.96 .006 Roles 2.3 2.58 0.37 1.75–3.38 2.45 0.37 1.75–3.38 1.57 .128 Affective Resp . 2.2 2.34 0.68 1.00–4.00 2.12 0.62 1.00–3.67 1.48 .151 Affective Inv. 2.1 2.37 0.28 1.71–3.00 2.44 0.23 1.86–3.00 - 1.61 .118 Behav. Control 1.9 2.12 0.44 1.00–3.11 1.77 0.41 1.00–2.78 4.23 < .001 Conflict --- 9.60 4.83 0 .00–18.00 10.05 5.67 1 .00–20.00 - 0.46 .649 Note. Cut-Off = McMaster Cut-Off score; Affective Resp. = Affective responsiveness; Affective Inv. = Affective involvement; Behav. Control = Behavioral control. We used SEM to test the fit of our theory-driven, congeneric model of family functioning using seven subscales from each source (14 variables; seven for adolescents and seven for caregivers, with the error terms of each subscale correlated between the two sources) to predict family functioning as reported by each source (two latent variables; one for adolescents and one for caregivers). The absolute fit of the model was marginal: χ 2 (69) = 104.39, p = .004, CFI = 0.79, RMSEA = 0.14, SRMR = 0.14. In order to reduce variables in our theoretical model, we averaged adolescent and caregiver reports for problem-solving, roles, affective responsiveness, affective involvement, and conflict because these did not statistically differ (all values of p > .05). However, we kept the two reports as separate predictors for communication and behavioral control. This left us with nine predictor variables for subsequent analysis (five averaged predictors and four single-source predictors). Next, we used SEM to test the fit of the simplified model with the nine observed variables and one latent variable of family functioning. We found that the absolute model fit of this simplified model was acceptable overall. Specifically, the fit indices mostly indicated good fit (χ2[27] = 33.11, p = .194, CFI = 0.93, SRMR = 0.08), though one fit index suggested poor fit (RMSEA = 0.09). Differences in the chi-squares of our two models showed the simplified model was statistically better than the initial model: χ 2 (42) = 71.28, p = .003. Thus, we selected the simplified model as the final model of family functioning (see Figure 1). See Table 2 for descriptive analyses of the nine predictors in the final model. All variables were positively related to family functioning. The strongest predictors of family functioning in this model were affective responsiveness (average of adolescent and caregiver report; β = .85, B = 1.04, SE B = 0.21, p < .001, R 2 = .72), affective involvement (averaged; β = .72, B = 0.88, SE B = 0.22, p < .001, R 2 = .51), and problem-solving (averaged; β = .82, R 2 = .67; this was the constrained parameter used to identify the regression model).