Data Analytics: Foundations to Practice · Correlation, Relationships, and Causation
What Correlation Actually Measures
Correlation is one of the most useful and most commonly misused concepts in data analytics, largely because its precise meaning is narrower than how it's typically talked about. This chapter pins down exactly what it measures.
Correlation measures the extent to which two variables tend to move together, described by both direction and strength: a positive correlation means both variables tend to increase together, a negative correlation means one tends to increase as the other decreases, and the strength of a correlation, commonly expressed as a correlation coefficient ranging from negative 1 to positive 1, indicates how closely and consistently the two variables actually track each other.
- Correlation measures both direction (positive or negative) and strength (via a coefficient from -1 to +1) of how two variables move together.
- A coefficient near ±1 indicates a strong relationship, near zero indicates little linear relationship, with no single universal threshold for 'strong' across all fields.
- The standard correlation coefficient only captures linear relationships; a strong non-linear pattern can show a near-zero coefficient despite being genuinely predictable.
- Visualizing a relationship via scatter plot remains important, since a single coefficient can hide non-linear patterns or a relationship driven by just one or two outliers.