Table 2.
HLM Results for Each Outcome Measure
| AAQ | DBT Cohort 1 | −1.95 | 0.42 | - | 92.9 | < | Unstructured, Autoregressive |
| DBT Cohort 2 | −3.61 | 0.42 | - | 67.1 | < | ||
| CTBE Cohort 1 | −0.86 | 0.45 | - | 90.7 | .06 | ||
| CTBE Cohort 2 | −2.51 | 0.45 | - | 71.9 | < | ||
| DBT vs. CTBE | −1.09 | 0.50 | - | 79.9 | .03 | ||
| Cohort 1 vs. 2 | 1.65 | 0.50 | 3.3 | 79.8 | < | ||
| STAXI Anger Control | DBT | 0.04 | 0.01 | 86.1 | < | Unstructured | |
| CTBE | 0.02 | 0.01 | 92.1 | .23 | |||
| DBT vs. CTBE | 0.02 | 0.02 | 89.4 | .21 | |||
| STAXI | DBT | −0.09 | 0.01 | - | 79.5 | < | Unstructured, Autoregressive |
| Anger In | CTBE | −0.06 | 0.01 | - | 83.0 | < | |
| DBT vs. CTBE | −0.03 | 0.02 | - | 81.5 | .15 | ||
| STAXI Anger Out | DBT | −0.05 | 0.01 | - | 88.1 | <.001 | Unstructured |
| CTBE | −0.01 | 0.01 | - | 96.6 | .37 | ||
| DBT vs. CTBE | −0.04 | 0.02 | - | 92.8 | .03 | ||
| STAXI Anger Trait | DBT Drops | −0.09 | 0.02 | - | 98.6 | < .001 | Unstructured, Autoregressive |
| DBT Completers | −0.05 | 0.01 | - | 81.2 | < .001 | ||
| CTBE Drops | −0.08 | 0.02 | - | 106. | < .001 | ||
| CTBE Completers | −0.03 | 0.01 | - | 80.4 | .02 | ||
| DBT vs. CTBE | −0.02 | 0.02 | - | 83.7 | .40 | ||
| Completers vs. Drops | 0.05 | 0.02 | 2.1 | 99.7 | .03 | ||
| PFQ Shame | DBT | −0.07 | 0.02 | - | 54.2 | < .005 | Unstructured, Autoregressive |
| CTBE | −0.08 | 0.02 | - | 56.6 | < | ||
| DBT vs. CTBE | 0.01 | 0.03 | 55.6 | .64 | |||
| PFQ Guilt | DBT | −0.07 | 0.03 | - | 51.2 | <.01 | Unstructured, Autoregressive |
| CTBE | −0.07 | 0.03 | - | 52.6 | .01 | ||
| DBT vs. CTBE | −0.001 | 0.04 | - | 52.0 | .98 | ||
| TMAS | DBT | −0.03 | 0.01 | - | 47.8 | < | Unstructured |
| CTBE | −0.02 | 0.01 | - | 47.7 | .01 | ||
| DBT vs. CTBE | −0.01 | 0.01 | - | 47.8 | .60 | ||
Note.
Each slope is compared to a null slope using a t-test, and the t value, degrees of freedom and significance are presented in this table.
A random intercept and random slope HLM model with an unstructured covariance structure for the random effects provided the best fit for the data for all outcomes. To test whether a model accounting for within subject correlation between time-points better explains the data, we fit an autoregressive (AR1) correlated residual error structure to each model (Gibbons et al., 1993). This more complex model was selected only if the model fit was significantly improved.
S.E.=standard error; AAQ= Acceptance and Action Questionnaire, STAXI= State Trait Anger Inventory, PFQ= Personal Feelings Questionnaire, TMAS= Tailor Manifest Anxiety Inventory, DBT=Dialectical Behavioral Therapy, CTBE=Community Treatment by Experts, Drops = participants who dropped out of treatment, Completers = participants who completed treatment.