Quick answer: In jamovi: Regression → Correlation Matrix measures the strength and direction of a relationship (Pearson r for linear numeric data, Spearman for ordinal or non-normal), while Regression → Linear Regression goes further and models the line itself so you can predict one variable from another. Correlation asks how tightly do they move together?; regression asks what line describes it, and how well? Both report p-values — interpreted exactly as in our p-value explainer.
Correlation first: the r you can defend
| r value | Reading |
|---|---|
| 0.0 – 0.3 (or −0.3 – 0.0) | Negligible to weak |
| 0.3 – 0.5 | Moderate |
| 0.5 – 0.7 | Strong |
| 0.7 – 1.0 | Very strong — check for near-duplicate variables |
Setup: Regression → Correlation Matrix → variables in, tick Pearson under Correlation Coefficients and the significance + CI options. Two rules reviewers enforce: Pearson demands roughly linear, normally-paired data (switch to Spearman when not), and r is dimensionless — height in cm vs weight in kg correlates identically to metres vs pounds. In a worked example, study-hours vs exam score gives r = 0.72, p < .001: a strong positive association, and — the sentence every methods chapter needs — correlation alone never proves cause.
Regression: from relationship to prediction
Regression → Linear Regression: move exam score to Dependent Variable, study-hours to Covariates. The output that matters: the slope (each extra study-hour predicts +6.1 points), the intercept, R² (0.52 — the model explains 52% of score variance), and the overall F-test with p < .001. Report as: “Study hours predicted exam scores, b = 6.1, R² = .52, F(1, 38) = 41.2, p < .001.” Tick Confidence interval under Model Coefficients and jamovi prints the CI for every coefficient; the residual plots under Assumption Checks (Q-Q, residual vs fitted) are the normality/equal-variance eyeball.
The R Commander route to the same model is mapped in our R Commander tests guide, and the earlier stepping stone — measuring groups rather than lines — is the t-test workflow. For one-to-one coaching on model choice, assumptions and reporting, Ampersand Academy teaches statistics with jamovi one-to-one.
Frequently asked questions
What is the difference between correlation and regression?
Correlation measures how strongly two variables move together with a single r between -1 and 1. Regression fits a predictive line and quantifies how much the outcome changes per unit of the predictor, plus how much variance the model explains.
When should I use Spearman instead of Pearson correlation?
When data are ordinal, clearly non-normal, or the relationship is monotonic but not straight-line. Spearman ranks the data first, so it tolerates skew that would distort Pearson.
What does R-squared mean in jamovi regression output?
R-squared is the share of outcome variance the model explains. An R-squared of 0.52 means 52 percent – but a high value never proves causation, only predictive fit on the data at hand.
How do I read the slope coefficient?
The slope is the predicted change in the outcome per one-unit increase in the predictor, holding other predictors constant. Always report it with its confidence interval from the Model Coefficients table.
Can correlation be significant but weak?
Yes – with large samples even r = 0.15 turns significant. That is why effect size and practical meaning matter alongside the p-value; the p-value only rules out chance.