6: Foundations of Design Flashcards
Non-experimental and experimental designs are classified as what two things each?
Non-experimental: descriptive; correlational.
Experimental: non-randomised (quasi-experimental); randomised.
Other term for base rate, regarding disorders?
Prevalence.
Studies that look at relationships between variables, but do not experimentally manipulate anything, are what?
Correlational designs.
What two correlational designs can you take?
Cross-Sectional: all observations made only once at a single time.
Longitudinal: measurements made at 2 or more time points.
What four elements do you need to infer causality from correlational research? Define each.
Covariation: two variables must occur together.
Precedence: hypothesized causal variable must reliably precede the outcome variable.
Exclusion of alternative explanations: other explanations for observed covariation must be reasonably excluded.
Logical mechanism: must be plausible account (i.e., THEORY) for the hypothesized causal relation.
What are two cautions that must be taken when interpreting correlational research?
Bidirectionality: two-way causality.
Spurious association: not genuine; 3rd variable problem.
A cross-sectional correlation coefficient CANNOT determine _____, but can partially deal with the _____ if control variables are used.
Directionality; 3rd variable problem.
What are two primary 3rd variable problems?
Mediation and moderation.
What are two quasi-experimental designs?
One-group posttest-only design.
One group pretest-posttest design.
What are two experimental designs? Under what conditions are they quasi-experimental?
Posttest only; pretest-posttest.
Quasi-experimental if groups are not randomized.
Both experimental designs have at least two what?
Groups.
Which trial is the gold standard in clinical psychology?
Randomised Controlled/Clinical Trial.
What are the five types of control group?
No-treatment controls.
Wait-list controls: delay in treatment.
Placebo controls: credible but inert treatment.
Comparative treatment groups: alternative treatment which is also effective.
Dismantling studies: break treatment into components, use each component in isolation as a group.
List eight good experimental design features.
Patient homogeneity.
Randomised assignment.
Specific interventions.
Appropriate control groups.
Groups treated equivalently except for intervention.
Low attrition.
Patients, clinicians and raters blind.
Independent replication.
Define external and internal validity.
External: to what extent can study results be generalized to other samples with different characteristics than the study sample?
Internal: degree to which causality can be inferred from a study.
Internal validity measures the extent to which what factors and not others account for study results?
Intervention, manipulation and/or experimental procedures.
List three threats to internal validity.
Spontaneous remission: recovery from disorder without any apparent external reason or intervention.
Interfering events (individual) and Secular Drift (long-term social changes): should be applicable to most of sample.
Maturational trends: relevant to treatment outcome studies and children.
What is regression to mean? When is it problematic?
Tendency for extreme scores to revert, or regress, toward mean of distribution when measurement is re-administered.
When groups are initially selected on the basis of extreme scores.
What is attrition? When is it a problem for internal validity?
Loss of participants over time.
Different rates of attrition between conditions; different reasons for attrition between conditions.
What two disorders are associated with the highest rates of attrition?
Depressive disorder.
Substance use disorder.
Some participants in a ‘control’ group may receive aspects of the intervention provided to the ‘experimental’ condition. What is this called?
Diffusion/imitation of treatment.
How is special treatment/reaction of controls a potential threat to internal validity?
‘Control’ subjects may be aware that they are not receiving treatment and be motivated to ‘out-perform’ experimental group.
May feel demoralized or lose interest/motivation in participation if no active treatment is given.
Relating to internal validity - if one doesn’t attend all therapy sessions or doesn’t take medication as prescribed, they have what?
Poor adherence to treatment protocol.
Larco et al. (2002) found that adherence to taking anti-psychotic medication among people with schizophrenia was around 50%. What three elements worsened adherence?
Poor insight, negative feelings towards medication, poor therapeutic alliance.
What are three threats to construct validity of experimental designs?
Confounded manipulation: sometimes accidentally manipulate more than one thing at a time.
Expectancy effects: sometimes people get better simply by thinking the therapy will work.
Hawthorne effect: sometimes people appear to get better simply because they’re being observed.
Define statistical conclusion validity.
Extent to which the analyses performed enables one to draw correct inferences about the phenomena of interest.
List five threats to statistical conclusion validity.
Low power: small N, small effect size.
Improper data analysis: e.g., analyzing everything and reporting the “best” results.
Confounding variables.
Poor reliability of measures.
Poor validity of measures.
In statistical analyses, as reliability decreases, _____ increases.
Type II error.
The Dead Salmon Study brought light to what problem?
Multiple comparisons problem: if you do many different statistical tests, some of them will give interesting results by chance.
The most prestigious journals favor what two things?
Novel results.
Statistically significant results (p < .05). Null findings suppressed, leading to “file drawer effect.”
