Correlations

A correlation looks for a significant relationship between two variables.

E.g. does memory increase change with increasing age?

Values range from -1 to +1.

Hypothesis: Positive or negative relationship?

Results: There is a significant positive relationship between emotional IQ and the ability to recognise facial emotions (r = .660, p = .002).

If significant then create a scatterplot, and include a line of best-fit.

Residual = difference between raw data point and best-fit line.

CORRELATION DOES NOT IMPLY CAUSATION

There could be a 3rd variable affecting the results – we can control for this.

E.g. there are more ice cream sales and more shark attacks: when it is hot more people buy ice creams and more people swim in the sea.

With facial emotion recognition and emotional IQ, a potential confound could be social anxiety.

We can control for temperature when analysing the ice cream sales and shark attacks correlations.

The confound may explain some of the random (unexplained residual) variance.

Partial correlation: Is there a correlation after controlling for social anxiety score.

Results: After controlling for social anxiety, there is a positive correlation between the recognition of facial emotion and emotional intelligence (r = .467, p = .044).

Does the strength of the correlation change from the original to the partial? If it weakens, then it tells us that some of the variance can be explained by social anxiety.

Sex could also be a confound: Is the correlation between emotional IQ and emotion recognition significantly different for males and females?

You could compare these correlations through splitting file for sex: Males = no correlation, females = positive correlation. There is a sex difference in the relationship between accuracy for emotional face processing and emotional intelligence.

You can graph these results through two correlations, and test whether the correlations are significantly different (need to go to a different website to do this).

You get a z value, which tells you whether the correlation is significantly larger in males than females.

	Created by PPKD over 10 years ago

Correlations, partial correlations, confound, unexplained random variance, split file, z value