Lecture 16 - Correlation

Question 1

What is a correlation?

Answer 1

• don’t manipulate any variables
• measure what is naturally occurring between different variables
• does not imply causation

Question 2

What graph do you use to represent a correlation?

Answer 2

• scatter plot
• should use a line of best fit = illustrates the overall relationship between your 2 variables across all of your participants

Question 3

What does a positive correlation mean?

Answer 3

as scores on 1 variable increase, scores on the other variable also increase

Question 4

What does a negative correlation mean?

Answer 4

as scores on 1 variable increase, scores on the other variable decrease

Question 5

What does no relationship mean?

Answer 5

as scores on 1 variable increase, scores on the other variable change but in a random way

Question 6

What kind of variables should you use in a correlational study?

Answer 6

• no IV or DV
• both variables must give you a continuous score
• cannot calculate a correlation from variables that are categorical

Question 7

Pearson’s R correlation?

Answer 7

• we end up with an R value = maps directly onto the type of relationship that exists within your data sets
• the R varies from -1 through to +1
• a value of -1 =perfect negative relationship, +1=perfect positive relationship between variables, 0 =no relationship

Question 8

What does a Pearson’s correlation do?

Answer 8

• looks at how much the data points change together which is known as covariance
• if all the data points are very close to the line of best fit then the relationship is very consistent whereas if they are far away from this line there will be more variability in this relationship
• also looks at how much the data point change together which is known as Dooku variance
• R = covariance between the variables divided by variance within the variables

Question 9

Which variables can’t you use for Pearson’s?

Answer 9

• cannot use categorical or nominal
• can only use ratio or interval as it is a parametric test
• is sensitive to outliers so must not have any strong outliers

Question 10

Spearman’s rho?

Answer 10

• non parametric
• use if you have ordinal data
• looks for a relationship between 2 continuous variables however one or both of them will violate the assumption that one of the variables may have not been measured at a ratio level
• isn’t as sensitive to outliers