Outcome variance — the degree to which therapeutic results differ across individuals, programs, and measurement occasions — occupies a recurring methodological and interpretive position across the depth-psychology and behavioural-intervention corpus. The literature does not treat outcome variance as a nuisance to be eliminated but, rather, as a signal demanding explanation: heterogeneity statistics (I², Q-values) in meta-analytic work by Bettmann and Bowen reveal that the true variance among effect sizes substantially exceeds what sampling error alone can account for, compelling the search for moderators. Russell's programme-evaluation studies frame variance questions developmentally, asking how outcomes differ by age, gender, diagnosis, and programme length. Benda's structural-equation analyses partition unique variance attributable to religiousness, spirituality, and social attachment, distinguishing direct from indirect pathways to substance-use outcomes. DeMille employs regression modelling to isolate the fraction of post-treatment variance explicable by treatment participation versus intake severity. Across these traditions, a central tension persists: whether residual, unexplained outcome variance reflects unmeasured individual difference variables (readiness to change, attachment style, coping resources) or structural programme features (duration, therapeutic mode, staffing). The practical stakes are considerable — understanding what drives outcome variance informs both programme design and the equitable allocation of intensive interventions.
In the library
18 passages
the Q-value was higher than the degrees of freedom and therefore the null hypothesis that all studies shared the same true effect could be rejected. Thus, the conclusion could be drawn that the true effect size varied from study to study.
Bettmann establishes that outcome variance across wilderness therapy studies is genuine true variance rather than artefact, justifying moderator analysis across all six meta-analytic outcome domains.
Bettmann, Joanna Ellen, A Meta-analysis of Wilderness Therapy Outcomes for Private Pay Clients, 2016thesis
The I2-value of 65.32 indicates that 65.32 % of the variance is true variance in the studies rather than sampling error… Program model (Qbetween = 14.29, df = 1, p<.01) was found to be significant.
For personal effectiveness outcomes, Bettmann demonstrates that programme model accounts for a significant portion of between-study outcome variance, identifying a structural moderator.
Bettmann, Joanna Ellen, A Meta-analysis of Wilderness Therapy Outcomes for Private Pay Clients, 2016thesis
The I2-value of 85.85 indicates that 85.85 % of the variance is true variance in the studies rather than sampling error… Program model… use of a mental health practitioner… and publication year… were all found
Bettmann identifies programme model, mental health practitioner involvement, and publication year as significant moderators of interpersonal outcome variance in wilderness therapy research.
Bettmann, Joanna Ellen, A Meta-analysis of Wilderness Therapy Outcomes for Private Pay Clients, 2016thesis
variation (heterogeneity) among outcomes is related to particular characteristics of the studies. The dependent variable is the effect size and the independent variables (predictors) represent sample, program, and participant characteristics.
Bowen and Neill articulate the formal logic for treating outcome variance as a function of sample, programme, and participant predictors, employing weighted generalised least squares regression within a random-effects framework.
Bowen, Daniel J., A Meta-Analysis of Adventure Therapy Outcomes and Moderators, 2013thesis
this regression model accounted for 29.4% of the variances in Time 2 Y-OQ 2.01 total scores, suggesting that almost one third of the variance in Time 2 scores can be accounted for by this model.
DeMille quantifies the proportion of post-treatment outcome variance explained by OBH participation and intake severity, framing unexplained variance as a target for further investigation.
DeMille, Steven, The effectiveness of outdoor behavioral healthcare with struggling adolescents: A comparison group study, 2018thesis
How did treatment outcome vary according to age, gender, and DSM-IV diagnosis? How did treatment outcome vary according to program length? To what extent did clients maintain outcomes 12-months posttreatment?
Russell systematically operationalises outcome variance as a function of demographic, diagnostic, and programme-structural variables in outdoor behavioural healthcare, setting the agenda for person-by-programme interaction research.
Russell, Keith C., An Assessment of Outcomes in Outdoor Behavioral Healthcare Treatment, 2003supporting
Differences in outcomes were found across program models which were driven by program length… significant differences in discharge scores were noted, with shorter programs reporting higher discharge scores.
Russell attributes a meaningful portion of between-programme outcome variance to programme duration, while noting that long-term follow-up attenuates these differences.
Russell, Keith C., An Assessment of Outcomes in Outdoor Behavioral Healthcare Treatment, 2003supporting
The amount of unique variance explained by religiousness is 2.5 percent in alcohol use, 3 percent in use of other drugs, and 3.4 percent in delinquency.
Benda partitions unique outcome variance attributable to religiousness versus church attendance, demonstrating that spirituality accounts for a small but consistent increment in substance-use outcome prediction.
Benda, Brent B., Spirituality and Religiousness and Alcohol/Other Drug Problems: Treatment and Recovery Perspectives, 2006supporting
R2 is the amount of variance accounted for by predictors… Alcohol/drug abuse (R2=.44) Spiritual well-being… Self-development… Abuse… Distress… Depression… Relation problems.
Benda's path model accounts for 44% of the variance in alcohol and drug abuse outcomes, disaggregating contributions of spiritual well-being, self-development, abuse history, and relational distress.
Benda, Brent B., Spirituality and Religiousness and Alcohol/Other Drug Problems: Treatment and Recovery Perspectives, 2006supporting
the full model was significant, accounting for 21% of the variance in F1 QOL, as was the change in R2 resulting from adding recovery capital to baseline QOL level (a 10% increase).
Laudet demonstrates that recovery capital variables explain a significant and incremental portion of quality-of-life outcome variance beyond baseline functioning, with spirituality and stress as the dominant predictors.
Laudet, Alexandre B., Recovery Capital as Prospective Predictor of Sustained Recovery, Life Satisfaction, and Stress Among Former Poly-Substance Users, 2008supporting
The GSE score at t0 emerged as a significant predictor (β=0.69, p<0.001) and explained 35.2% of the variance… none of the newly added variables represented a significant predictor because only an additional 1.6% of the variance could be explained.
Kratzer's hierarchical regression shows that baseline self-efficacy dominates outcome variance in bouldering psychotherapy, with demographic and treatment variables explaining negligible additional variance.
Kratzer, André, Bouldering psychotherapy is effective in enhancing perceived self-efficacy in people with depression: results from a multicenter randomized controlled trial, 2021supporting
Unique % variance explained: religious coping… Event-Specific Outcome 11%… Positive Affectivity 6%… Negative Affectivity 7%… General Health Questionnaire 9%
Pargament's multi-study table systematically documents the unique variance in adjustment outcomes attributable to religious coping over and above generalised religiosity and demographic controls.
Pargament, Kenneth I, The psychology of religion and coping theory, research,, 2001supporting
time in treatment was significantly and negatively related to OQ scores (β = −2.11, p = .024). The standardized coefficient
Russell's hierarchical linear modelling isolates within-client and between-client sources of outcome variance, identifying treatment duration and ATES process factors as significant predictors of OQ-45 change.
Russell, Keith C., Process Factors Explaining Psycho-Social Outcomes in Adventure Therapy: The Adventure Therapy Experience Scale (ATES), 2017supporting
Bowen and Neill (2013) in their meta-analysis of 197 studies on adventure therapy… found age to be a significant predictor of outcomes with older participants reporting larger improvements than younger participants.
DeMille corroborates meta-analytic findings that participant age is a reliable source of outcome variance in adventure and outdoor behavioural healthcare, with older adolescents showing greater gains.
DeMille, Steven, The effectiveness of outdoor behavioral healthcare with struggling adolescents: A comparison group study, 2018supporting
a random effects model was used to account for heterogeneity across the studies and greater generalizability of the results… The Q-statistic provided information about the ratio of observed variance to the within-study error or heterogeneity of the study.
Bettmann articulates the methodological rationale for random-effects modelling as the appropriate framework when true outcome variance across studies is expected to exceed sampling error.
Bettmann, Joanna Ellen, A Meta-analysis of Wilderness Therapy Outcomes for Private Pay Clients, 2016supporting
the only predictors of meaningful change at Time 2 were that participants were experiencing some level of clinical distress at intake/inquiry and that they engaged in an Outdoor Behavioral Healthcare intervention.
DeMille finds that intake severity and treatment modality together account for the explained portion of outcome variance, with demographic variables contributing no independent predictive power.
DeMille, Steven, The effectiveness of outdoor behavioral healthcare with struggling adolescents: A comparison group study, 2018supporting
62.3% were negative (143 rated as −2, 192 rated as −1) and 37.7% were positive (56 rated as +1, 147 rated as +2). A total of 183 studies (53% of those from which an OLS could be obtained) yielded at least one positive contrast showing a treatment effect.
Miller's outcome logic scoring framework captures the distribution of treatment effects across alcohol use disorder trials, providing an implicit map of outcome variance across modalities without directly modelling its sources.
Miller, William R., Mesa Grande: a methodological analysis of clinical trials of treatments for alcohol use disorders, 2002aside
Because of the wide range of programmes using NAT for a wide range of outcome goals it shou
Annerstedt identifies heterogeneity of outcome goals across nature-assisted therapy programmes as a structural reason why outcome variance is difficult to interpret without typological classification of therapeutic intent.
Annerstedt, Matilda, Nature-assisted therapy: Systematic review of controlled and observational studies, 2011aside