Correlations (1)

What is a correlation ?

A correlation is not a research method as such, rather it is an analysis of the relationship between co - variables

Are the variables manipulated in correlational research ?

• In correlational research, the variables are not manipulated (there is no IV), instead two co-variables are measured and compared to look for a relationship

• One or both of the co-variables could be pre-existing e.g.

  • school attendance measured as days present at school in Year 11 (co-variable) and number of GCSEs achieved (co-variable)

  • average temperature in August (co-variable) and number of arrests made for violent behaviour in August in one town
    

• One or both of the co-variables could be measured for the research itself e.g.

  • number of arguments with your partner in a month (co-variable) and self-reported stress levels for the same month (co-variable)

  • average number of hours sleep in one week (co-variable) andnumber of cups of caffeinated beverages consumed in the same week (co-variable)

What scores does a correlation use ?

• A correlation uses two scores e.g. number of cups of caffeine and the number of hours of sleep

• In the case of self-reported data, there are two scores per participant e.g.

  • an average of 4 hours of sleep per night correlated with 58 cups of caffeine consumed in a week

• In the case of pre-existing data, the researcher would simply go to the records e.g.

  • Student X was present for 188 days in Year 11 - they achieved 10 GCSEs 

How many scores does each participant have ?

• Each participant has two scores which the researcher then calculates to look for a relationship e.g.

  • is there a relationship between the number of cups of caffeine consumed in one week and the number of hours of sleep achieved in the same week?

How are scores for one co - variable and the score for the other co - variable plotted ?

• The score for one co-variable and the score for the other co-variable are plotted as one point (usually represented as an 'x') on a scattergraph

What outcomes do scattergraphs show ?

• Scattergraphs show one of three outcomes 

  • Positive correlation

    • One co-variable increases as the other increases (but not necessarily at the same rate) e.g. calories consumed and weight gained

  • Negative correlation

    • One co-variable increases while the other co-variable decreases  (but not necessarily at the same rate) e.g. hours spent sitting down and level of fitness

  • Zero correlation

    • There is no relationship between the co-variables e.g. hair colour and IQ

How can we analyse the relationship between co - variables ?

• Analysing the relationship between co-variables can be done by

  • visually 'eyeballing' the scattergraph to see the direction of the relationship (positive, negative or none at all)

  • calculating the correlation coefficient which is expressed as a numerical value

How to calculate the correlation coefficient ?

• The correlation coefficient represents both the direction and thestrength of the relationship between the co-variables, expressed as a value between -1 and +1

  • A perfect positive correlation would be expressed as  +1

  • A perfect negative correlation would be expressed as  -1 

  • No relationship would be expressed as 0 

What can be described as weak , moderate or strong ?

• Both positive and negative coefficient correlations can be described as weak, moderate or strong e.g.

  • a correlation coefficient of 0.3 is a weak positive correlation

  • a correlation coefficient of -0.9 is a strong negative correlation

  • a correlation coefficient of -0.5 is a moderate negativecorrelation

Evaluation of types of correlation - Strength

• The data may be easily available for researchers to quickly analyse 

  • This is a strength as it enables the researcher to access large amounts of data that would otherwise be impossible to gather if they tried to amass this from scratch

  • Large amounts of quantitative data mean that the research is high in reliability

• Correlations allow researchers to make predictions as to the relationship between co-variables e.g.

  • Knowing that there is a relationship between school absence and GCSE results could be used to identify students at risk and to implement interventions to help them achieve their potential

Evaluation of types of correlation - Limitations

• Extraneous factors connected to one or both co-variables may affect the result and lead to invalid conclusions being made e.g.

  • number of days of absence from school may be due to illnessrather than to choice 

  • a low GCSE score may be due to a high turnover of teachers in one school rather than to student absence

• Correlations work well for linear relationships e.g. height and shoe size

  • They are less successful when dealing with non-linear relationships e.g. number of hours worked and level of happiness

    • This limits the type of data that can be analysed and conclusion drawn

Experiment and correlation difference

• Experiments involve deliberate manipulation an independent variable (IV) to measure the impact on the dependent variable (DV)

  • This means that the researcher works towards establishing cause-effect e.g.

    • does a 30-second distraction task (IV) affect the number of items recalled from a list (DV)?

• An experiment measures the difference in participant performancedepending on how the researcher has manipulated the IV (condition 1 compared to condition 2)

Experiment and correlation difference

• With a correlation, there is no researcher manipulation of either co-variable

• Both co-variables may have a bi - directional influence on each other

  • A correlation measures the strength and direction of a relationship between co-variables compared to an experimentwhich measures a difference in conditions

• Correlations cannot establish cause-effect, even when results show a strong positive correlation

  • One variable may impact the other but not in a linear fashione.g.

    • number of arguments with your partner in one month may lead to higher stress levels but higher stress levels are also likely to lead to more arguments

  • Thus, it is difficult to know which (if any) of the two co-variables initially triggered the other or if some other factor is the reason for the arguments/stress