Correlation coefficient is one way to try to measure if two different things have a connection, for example daily temperature and ice-cream sales.

If the coefficient's value is 1 it means the two things move totally together.

It it's -1, it means the two things move in exactly the opposite way to each other, so if one goes up, the other goes down in proportion.

When it's 0, it means the two things don't move together at all, there is no apparent connection between them.

Most of the time, values aren't anything like 0, 1 or -1. Typically a value of 0.7 would be thought of as a high degree of correlation.

There are two important things to remember about correlation coefficients.

1) It is typically only looking for a certain simple ("linear") relationship between the two things being compared. There might be a strong relationship, but not of the simple kind that correlation coefficients are designed to measure.

2) There is a saying: "Correlation does not equal causality". That means just because two things are highly correlated, it doesn't mean one *causes* the other. For example if we found a strong correlation between drinking red wine and liking garlic, it doesn't mean that drinking wine *makes* you like garlic, or that liking garlic causes you to drink red wine.

Further reading:

http://www.socialresearchmethods.net/kb/statcorr.php

Permalink | Report