Key Points: Research Design, Sampling, and Correlation vs Causation
Operationalization and Study Design
- Outcomes must be operationalized: exactly how a variable is measured (e.g., happiness, success).
- Clear design guides what you can claim from findings.
Sampling and Representativeness
- Much psychology uses college students; often WEIRD (Western, Educated, Industrialized, Rich, Democratic).
- College samples may not reflect all ages, socioeconomic statuses, or country contexts.
- College students differ from non-students in resources, opportunities, and background; limits external validity.
- Practical reasons (funding, access) drive participant pools; community outreach can broaden samples but is more work.
- Large, diverse samples help balance confounds; measuring relevant covariates helps control for differences.
Population Generalization
- Findings may not generalize beyond the studied population or setting (e.g., front-row effects in large lectures vs. labs).
- Replication across ages, settings, and populations strengthens claims.
Data Types: Quantitative vs. Qualitative
- Quantitative: numerical data (scores, times, scales); analyzed with statistics.
- Qualitative: non-numerical data (interviews, diaries); analyzed for themes.
- Both approaches complement each other; mixed-methods are common.
Descriptive Statistics
- Mean: average; sensitive to outliers.
- Median: middle value; robust to outliers.
- Mode: most frequent value; useful for bimodal data.
- When outliers exist, consider median or mode to understand central tendency.
Inferential Statistics
- Descriptive statistics describe the sample; inferential statistics generalize to populations.
- Power and sample size affect detection of real effects; larger samples can yield statistical significance for small differences.
- Variance (spread) matters for determining significance.
Correlation vs Causation
- Correlational studies: observe natural behavior without manipulation; measure relationships between two variables.
- Correlation does not imply causation: third variables or reverse causation may explain relationships.
- Example: sitting in front of class correlates with higher grades, but not proven to cause higher grades.
Direction and Strength of Correlations
- Direction: positive (both rise) or negative (one rises, the other falls).
- Strength (rough guide):
- Weak: ∣r∣<0.3
- Moderate: 0.3≤∣r∣<0.6
- Strong: ∣r∣≥0.6
- Example values:
- Negative, weak: r≈−0.37
- Positive, strong: r≈0.86
- No correlation: r≈0.04
Illustrative Correlational Example
- Front-row vs final grade (correlational only): higher front seating associated with higher grades but causality not established.
- Random assignment would be needed to infer causality.
When Correlational Research is Valuable
- Helps detect relationships, direction, and strength before conducting experiments.
- Ethically or practically constrained questions (e.g., smoking) may rely on correlational data.
Experimental Design and Causality
- Experimental study: random assignment to conditions; includes a control group.
- If a significant effect is found, you can infer a causal relationship with reasonable confidence (not 100%).
Ethical and Practical Considerations
- Some questions cannot be randomized for ethical reasons (e.g., smoking exposure, pregnancy outcomes).
- Triangulation across methods (correlational data, experiments, longitudinal studies) strengthens causal inferences.
Longitudinal Studies
- Track behavior and outcomes over long periods; more powerful for detecting lasting effects.
- Harder to maintain participation; longer time, higher cost.
Summary of Correlational Evidence and Causality
- Correlation reveals relationships; does not prove causation.
- Experiments with random assignment are the gold standard for causal claims.
- Large samples, randomization, and replication across contexts bolster causal interpretations.
Practical Takeaways
- Before acting on findings, check how variables were measured (operationalization).
- Assess representativeness and generalizability of the sample.
- Distinguish between correlation and causation; use experiments when causal claims are needed.
- Consider ethical constraints and use multiple methods to triangulate evidence.
- Range of correlation: r∈[−1,1]
- Power concept (rough): larger samples increase the chance to detect a true effect.
- Descriptive vs. inferential statistics: describe data vs. generalize to populations.
- Means, medians, modes as measures of central tendency to summarize data.
- No-claim rule: correlation does not imply causation; causality requires experimental manipulation or strong quasi-experimental evidence.