# Correlation

## You are here

## Correlation

**Correlation** is a statistical technique used to **quantify** the **strength of relationship** between two variables.

Used a lot in psychology investigations, for example **Murstein (1972)** carried out a correlation analysis of ratings of attractiveness in partners ('computer dance' study).

Strengths: |
Weaknesses |
---|---|

Calculating the strength of a relationship between variables. |
Cannot assume cause and effect, strong correlation between variables may be misleading. |

Useful as a pointer for further, more detailed research. |
Lack of correlation may not mean there is no relationship, it could be non-linear. |

For a correlational study, the data can be plotted as points on a **scattergraph**. A line of best fit is then drawn through the points to show the trend of the data.

If both variables increase together, this is a **positive correlation.**

If one variable increases as other decreases this is a **negative correlation.**

If no line of best fit can be drawn, there is **no correlation.**

Correlation can be quantified by using a **correlation coefficient** - a mathematical measure of the degree of relatedness between sets of data.

Once calculated, a correlation coefficient will have a value from -1 to +1.

**+1 = perfect positive correlation** all points on straight line, as x increases y increases. A value close to one indicates a strong positive correlation.

**0 = no correlation** points show differing degrees of correlation.

* Note:* A correlation around zero may disguise a non-linear relationship.

**-1 = perfect negative correlation** all points on straight line, as x increases y decreases. A value close to -1 indicates a strong negative relationship.

* Note:* In real life human situations, or psychology experiments you will not find perfect correlation between variables, life is just like that.

What psychologists do is calculate a **correlation coefficient**, then, using statistical tables (thought up by brilliant mathematicians) work out the **probability** that their results could have occurred at random.

If they can say there is a 95% chance of their results really showing a strong correlation, then they accept that there is one.