Your starting point
Prerequisites: Use the ideas and vocabulary from lesson 17; revisit its worked examples if needed.
Definition: Correlation describes association. Pearson’s r summarizes the direction and strength of a linear relationship between paired numeric values, ranging from −1 to +1 when defined. Causation concerns whether changing one factor produces a change in another under relevant conditions.
Keep observations paired
Each x and y value must refer to the same unit or matched occasion. Sorting the columns separately can manufacture a relationship. Missing values should be removed as pairs for the correlation calculation, with the valid pair count reported. A constant variable makes Pearson’s r undefined because there is no variation to compare. Zero correlation does not rule out a nonlinear relationship.
Inspect the plot and possible explanations
A positive association can reflect a causal effect, reverse direction, a common cause or selection. Study hours and scores may both relate to prior knowledge or motivation. A single outlier can strongly influence a correlation, especially in a small sample. Plot the pairs, inspect groups and ask what design would distinguish explanations before attaching a causal verb.
Separate description from intervention
A correlation can be useful for description or prediction while being insufficient for deciding what would happen if you intervened. An experiment with appropriate random assignment can strengthen causal reasoning, but implementation, measurement and exclusions still matter. AI-generated wording often slips from “associated with” to “improves.” Check each causal claim against the actual design rather than the confidence of its phrasing.
Two worked examples
Perfect association, limited conclusion
Paired values x = 1,2,3 and y = 2,4,6 have Pearson r = 1. They lie on a straight rising line. This mathematical relationship alone says nothing about which variable influences the other. Both might simply be derived from the same record index in a synthetic example.
A common cause
Ice cream purchases and heat-related complaints may both increase during hot weather. Temperature is a plausible common cause. Concluding that buying ice cream causes complaints ignores that alternative. Measuring temperature and examining a suitable design would be more informative than merely calculating a larger correlation.
Where this becomes useful
A course analyst sees that learners who submit more practice score higher. The pattern may guide further investigation, but motivated or experienced learners could both practise more and score better. Do not promise a causal gain from the association alone.
Make an attempt
Use Correlation Investigation with the default pairs, then change one y value dramatically. Observe the scatterplot and r together. Try a constant x column and explain why the tool reports an undefined result rather than zero association.
Misconception check
Claim: A high correlation proves an intervention will work.
Inspect the correction
Association does not identify what would happen under a change. Confounding, reverse direction and selection remain possible.
Independent application
Create a small paired dataset with a visible association. Calculate r, inspect the scatterplot, identify one plausible confounder and write a descriptive conclusion that avoids causal language. Propose additional evidence that would make an intervention claim more credible.
Completion records practice and self-review. It is not certification of statistical competence.
