Correlations Flashcards
What is statistical covariance?
To what extent do variables co-vary?
A measure of how variables change together
Looking at raw relationship
When is a correlation done?
When unethical to do an experiment
What does a positive covariance mean?
Variables tend to increase and decrease in tandem
What does a negative covariance mean?
One variable tends to increase when other decreases
How do you find the measure of average cross-product deviations between two variables?
Find average (mean) of each set of scores
Subtract the average from each scores which gives up the deviation or how much each score differs from the average
Multiply deviations from two sets together = cross-product deviation
What does it mean if the measure of average cross-product deviations between two variables is positive?
When one variable scores above average, other tends to do the same
What does it mean if the measure of average cross-product deviations between two variables is negative?
When one scores above average, other tends to score below average
What is the rule of thumb for covariance?
If positive, x and y increase together
If negative, x goes up and y goes down
If 0, no relationship
What are the problems with correlations?
Depends on variance and scale of measurement used
What is Pearson’s correlation coefficient (r)?
Used for parametric (normally distributed) data
Tells how much 2 variables are related to each other
Standardises covariances to create correlation coefficient
Always returns a value between -1 and +1
What does it mean if Pearson’s correlation coefficient is +1?
When X goes up, Y goes up too
What does it mean if Pearson’s correlation coefficient is -1?
When X goes up, Y goes down
What does it mean if Pearson’s correlation coefficient is 0?
No (linear) relationship between X and Y
What is Spearman’s Rho?
Non-parametric version of standard correlation coefficient test (doesn’t require same assumptions as parametric tests - normal distribution, independence of sampling)
Often used with ordinal or categorical data or data that doesn’t generally meet parametric assumptions
Data is ranked then correlations calculated based on these ranks
Always returns a value between -1 and +1
What does it mean if Spearman’s Rho is +1?
When X goes up, Y goes up too
What does it mean if Spearman’s Rho is -1?
When X goes up, Y goes down
What does it mean if Spearman’s Rho is 0?
No (linear) relationship between X and Y
Does correlation equal causation?
NO
Another variable could be influencing the relationship
