Interpreting the Data
❦
Analysis produces output; interpretation converts it into meaning. A strong interpretation is accurate, linked to the research questions and theory, honest about uncertainty, and expressed in clear language.
Principles of Interpretation
1. Answer the research question for each objective; do not report everything.
2. Describe, then explain: first state what you found, then why it might be so.
3. Separate statistical from practical significance. A small effect may be significant in a large sample; a large effect may be non-significant in a small one. Report effect sizes and confidence intervals (Cumming, 2014; Wasserstein & Lazar, 2016).
4. Respect the design: correlational data show association, not causation.
5. Consider alternative explanations and the limits of the sample.
6. Link to theory and previous studies.
7. Be consistent across tables, text, and conclusions.
Interpreting Common Results
| Result | How to interpret |
|---|---|
| Mean and standard deviation | Level and variation of the variable; compare with the scale midpoint or benchmarks |
| Cronbach's alpha | Internal consistency of a scale |
| Correlation r | Direction and strength; p-value tells whether it is likely to be due to chance |
| t-test | Difference between groups; report means, t, df, p, and Cohen's d |
| ANOVA | Whether group means differ; post hoc tests show which; report F, df, p, and eta squared |
| Chi-square | Association between categorical variables; report χ², df, p, and Cramer's V |
| Regression R² | Proportion of variance explained (about 0.02 small, 0.13 medium, 0.26 large in Cohen's guide) |
| Regression coefficient B or β | Change in the outcome for a unit change in the predictor, holding others constant |
| p-value | The probability of data at least this extreme if the null hypothesis were true; it is not the probability that the hypothesis is true |
| Indirect effect (mediation) | The effect of X on Y through M; supported if the bootstrapped confidence interval excludes zero |
| Interaction (moderation) | The effect of X on Y differs by the level of W |
| SEM and PLS-SEM paths | Standardized path coefficient, t-value or p-value, f², and R² for each hypothesis |
Supporting or Rejecting Hypotheses
For each hypothesis, say: the test used, the result (estimate, test statistic, p-value or confidence interval), and the decision (supported or not supported). Do not write that a hypothesis is "proved". If a hypothesis is not supported, treat it as a finding and discuss why.
An Example Paragraph
"H1 predicted that work-life balance has a significant negative effect on turnover intention. The regression results show a significant negative effect (β = −.42, p < .001), so H1 is supported. The effect is moderate, and work-life balance explains about 18 percent of the variance in turnover intention (R² = .18)."
Interpreting Qualitative Findings
Interpret by moving from description to explanation: what participants said, how themes relate to each other, what they mean for the research question, and how they compare with theory. Use triangulation to compare sources, and be explicit about the role of the researcher.
Presenting Tables and Figures
Each table or figure should be understandable on its own: a number, a descriptive title, labeled rows and columns, units, notes, and a source. Refer to each in the text and explain what it shows without repeating all the numbers. Avoid misleading scales, and use charts only when they help.
Common Mistakes
- Repeating the table in words.
- Claiming causation from correlation.
- Ignoring non-significant findings.
- Describing a result as "highly significant" without effect sizes.
- Over-generalizing from a convenience sample.
CHAPTER 16