2.2 Samples & Variables in Abnormal Psychology Research

Course Context

• Multi-week module on research methods & design (each week = different focus)
• Current video ⮕ emphasis on samples & variables within abnormal psychology research

Sampling Approaches

• Case Study (Single–Participant) Designs
  • Rich, detailed life-span data → can link unique history to pathology
  • Core Limitation: non-generalizability
  • Ex: Childhood neglect/abandonment common in Borderline Personality Disorder (BPD) cases ➜ yet many neglected children do not develop BPD ⇒ other interacting factors must exist
• Group Designs
  • Capture averages/"norms" but never guarantee individual applicability
  • Classroom illustration: students often reject findings because “that’s not true for me,” misunderstanding that research speaks to probabilistic trends, not certainties
• Choosing Case vs Group
  • Guiding questions: prevalence, feasibility, ethical constraints
  • High-base-rate disorders (depression, anxiety): group studies practical (≈ $18\text{–}20\%$ lifetime prevalence ≈ $\frac{1}{5}$ people)
  • Low-base-rate disorders (Dissociative Identity Disorder): often force reliance on case studies

Clinical vs Non-Clinical Samples

• Non-patient psychopathology research = assessing everyday populations (e.g., college students) without manipulating IVs
  • Predominantly correlational; still informative for etiology hypotheses
  • College pools: cheap, large, course-credit incentives, not automatically “clinical”
  • Everyone scores somewhere on symptom scales ➜ can study dimensional variation (e.g., depressive mood vs neuroticism)
  • Foundational vs Translational link:
  • Non-clinical work = foundational (build basic mechanisms)
  • Later clinic trials = translational (apply to diagnosed populations)
• Pros
  • Accessibility, cost-effective, generates preliminary data for interventions
• Cons
  • Lacks disorder-specific nuances; manifestations may differ qualitatively in true clinical samples

Analog / Experimental Psychopathology Models

• Purpose: model abnormal processes under controlled lab conditions
• Examples
  • $CO_2$ inhalation challenge → induces panic-like symptoms in healthy participants (panic disorder analog)
  • Targeted brain lesions in rats → schizophrenia-like deficits; used to study high cigarette consumption & symptom relief
• Validity Checklist (to avoid erroneous generalization)
  1. Thorough description of the real disorder
  2. Reliable replication of target phenomena in lab
  3. Convergent validity: lab model & clinical disorder tap same mechanisms
  4. Discriminant validity: model does not reflect irrelevant processes

Core Research Designs

• Correlational / Observational

  • Measure naturally occurring variables; no manipulation
  • Empirical, objective, replicable ideals apply
  • Pearson’s $r$: range $-1 \le r \le +1$
  • Magnitude ≈ strength; sign ≈ direction
  • Rule of thumb: $|r| \approx .20$ (small), $|r| \approx .50$ (medium)
  • Classroom Graphs
  • Negative: ↑ beers → ↓ psych exam score
  • Zero: shoe size ↔ exam score
  • Positive: ↑ class attendance → ↑ exam score
  • Pros: discovery, ethical flexibility
  • Cons:
  • Directionality ambiguity (mindfulness ↔ depression?)
  • Third-variable risk (heat causes ice-cream & homicides)
  • Misuse: students label a study “correlational” yet write causal hypotheses → conceptual error

• Experimental

  • Independent Variable (IV) / Predictor manipulated ⇒ observe effect on Dependent Variable (DV) / Outcome
  • Random assignment (true experiment) minimizes confounds; lack thereof ⇒ quasi-experiment
  • Only design permitting causal inference
  • Example: Randomize participants to mindfulness vs self-compassion vs wait-list control; DV: BDI depression scores
  • Key Threats & Mitigations
  • Group Means vs Individuals: averages may not fit every client (“lost person” critique)
  • Internal vs External Validity: lab efficacy ≠ real-world effectiveness
  • Experimenter Expectancy (Rosenthal): solved via single/double-blind
  • Demand Characteristics: solved with filler/distractor tasks
  • Placebo / Nocebo effects
    • Antidepressant responders after 1 week ⇒ placebo (drug onset 4–8 weeks)
    • Japanese “poison-sumac” & House TV examples illustrate nocebo

Variable Basics & Operational Definitions

• IV / Predictor: presumed cause (manipulated or categorical)
• DV / Outcome: presumed effect; changes “depend” on IV
• Not all studies designate IV/DV (pure correlation) except by convention in regression
• Operational Definition = explicit statement of what & how we measure/manipulate
  • Well-being could be: Happiness Scale score, Life Satisfaction Index, Perceived Stress (inverse)
  • Depression examples:
  • Beck Depression Inventory (self-report)
  • Functional impairment (work-days missed, social quality)
  • Behavioral/physiological markers: slowed gait, drooped eyelids, speech latency

Ethical / Philosophical Implications

• Need to balance generalizability (group data) with individual care in clinical settings
• Lab controls must not overshadow ecological validity; researchers must question applicability to diverse cultures (e.g., gender identity & societal acceptance)
• Awareness of replication crisis across sciences → rigorous transparency & documentation mandatory

Numerical & Statistical References (LaTeX)

• Lifetime prevalence of mood/anxiety disorders: $18\text{–}20\% \; (\approx \frac{1}{5})$
• Pearson correlation thresholds: $|r| \approx .20$ (small), $|r| \approx .50$ (medium), $|r| = 1$ (perfect)

Key Takeaways

• Selecting sample type (case, group, clinical, non-clinical, analog) depends on prevalence, feasibility, and research goals
• Correlational studies reveal relationships but cannot prove causation; beware directionality & confounds
• True experiments with random assignment enable causal claims; quasi-experiments useful when manipulation impossible
• Robust research demands clear operational definitions, control of bias, and consideration of real-world applicability