Correlations

• Correlation example

  
• Correlations illustrate the nature of an association between two co-variables (the two measured variables in a correlational analysis).

  • They assess if there is a link/relationship between two things and can help us to understand:

  1. What the link is
  2. 2. How strong the links is

  • Plotted in a scattergram
  • Co-variables on each axis
  • Strength of correlation is indicated by a correlation coefficient ranging from -1-+1
• Types of correlation

  • Negative – as one variable increases the other decreases (r=-1)
  • No – no relationship (r=0)
  • Positive – as one variable increases so does the other (r= +1)
• It is also essential to know what a correlation coefficient is, when to use it, and how to calculate it:

  • WHAT: A number between -1.0 and +1.0 that tells us how closely the co-variables in a correlational analysis are related.
• It is also essential to know what a correlation coefficient is, when to use it, and how to calculate it:

  (1)

  • WHEN: The data is related (two scores from the same P) and interval (fixed units with equal distance between).
  • HOW: Pearson’s r (a parametric statistical test of correlation).
• It is also essential to know what a correlation coefficient is, when to use it, and how to calculate it:

  (2)

  • WHEN: The data is related (two scores from the same P) and ordinal (ranked/ordered but the difference between each item is not the same).
  • HOW: Spearman’s rho (a non-parametric statistical test of correlation).
• Difference between correlation and experiment:

  • Correlation is a non-experimental method

  Experiments:

  • Have an IV and DV
  • Allow us to infer causality e.g. ‘the change in the DV was brought about by the change in the IV.’

  Correlations

  • Don’t have any manipulation variables
  • Not possible to infer causality
  • You must consider third party intervening variables e.g. ice cream sales and murders
• Strengths of correlations:

  • Good preliminary tool – can be used to assess strength and direction of relationship before conducting experiments.
  • Could suggest ideas for future research
  • Economical
  • Quick and easy to carry out
  • No need for controlled environment
  • Can use secondary data which saves times and money
• Limitations of correlations

  • Can only tell us how variables are related but not why
  • Lack of experimental manipulation and therefore control
  • Cannot demonstrate cause and effect and therefore can’t be sure which variable is causing the other to change
  • Direction of causality difficult – i.e. caffeine and anxiety – positive correlation but which causes which?
• Limitations of correlations

  • Confounding/Third variable problem e.g. Anxiety/caffeine/high pressure job
  • Misuse of results
  • Issues above can lead to results being misused and misinterpreted
  • Relationships sometimes quoted as facts when they are not e.g. single parent households and crime – misquoted as a fact