What is correlation?
Correlation measures how consistently two variables move together. The most common measure, the Pearson correlation coefficient, runs from -1 to +1. A value near +1 means the two rise together, and a value near -1 means one falls as the other rises. A value near 0 means there is no straight-line relationship. Spearman’s rank correlation covers relationships that hold in direction without following a straight line, and outliers sway it less.
Correlation describes association only. It does not show which variable drives the other, or whether a third factor drives both. Squaring the correlation between a single input and an outcome gives the coefficient of determination (R-squared) of a simple linear regression.
Correlation in practice
A multi-site medical practice finds that missed appointments rise with the number of days between booking and the visit. That correlation is useful for prediction. A no-show model can use lead time as a feature, and the front desk can send extra reminders for visits booked far ahead. It does not prove that shorter lead times would cut no-shows, because patients who book far ahead may differ in other ways. Answering that question requires changing the booking rules at some clinics and comparing results.
What to watch for
- A coefficient near 0 can hide a strong curved relationship, so plot the data before trusting the number.
- A handful of outliers can create or erase a correlation on their own.
- Scanning hundreds of variable pairs will turn up strong correlations by chance alone.