Lecture 1 - Correlation Flashcards
What are 6 difficulties with correlations?
- Sample size
- Non-normal distributions
- Outliers
- Range effects
- Heterogenous samples
- Non-linear relationships
Why is sample size a difficulty with correlations?
Large sample sizes required to avoid errors, as small samples are very prone to them.
Why are non-normal distributions a difficulty with correlations?
If distributions are very far away from normal, the correlation co-efficient will not be helpful. A Pearson’s correlation can be used if the distribution is non-normal, but not too far away from normal.
Why are outliers a difficulty with correlations?
Single data points change the correlation co-efficient so that it does not reflect the rest/majority of the data.
Why are range effects a difficulty with correlations?
There may be an relationship for a specific section of data, but may not be outside of that section/outside of that range.
Why are heterogeneous samples a difficulty with correlations?
A scatter plot of two separate samples may, together, look as if there is a correlation, but a scatter plot of each separate sample may reveal that no correlation actually exists.
Why are non-linear relationships a difficulty with correlations?
There may be a systematic relationship despite a lack of linearity in the data - this means that it is more difficult to identify with Spearman’s and Pearson’s correlation.
Pearson’s r is a measure of what?
Effect size - it gives us an estimate of variance explained (r^2).
What are 4 reasons we want to carry out questionnaires?
- To test people
- To test new theories
- To understand behaviour
- To test change
What are acquiescent response sets?
Responses from participants that agree with anything that is said or suggested.
What are desirable response sets?
Responses from participants that answer according to what they think is desirable.
How might acquiescent or desirable response sets be accounted for?
- Invert some of the questions and then code them in reverse.
- Include contradictory statements/questions and see if participants agree with their own answers.
- Include dummy/masked questions.
- Try an alternative wording at the pilot stage.
How might acquiescent response sets be accounted fo and why?
Including contradictory statements - can’t say yes to both.
How might desirable response sets be accounted for and why?
Including dummy or masked questions - throws participants off the purpose of the questionnaire.
What does including contradictory statements help to account for?
Acquiescent response sets.
