Factor analysis Flashcards
What is factor analysis used for?
Although factor analysis is typically associated with questionnaire data, it can be used on any data set with lots of variables
How do you reduce all of your questions into a smaller number of more meaningful variables?
- Understand distinct aspects of a questionnaire
- Form a smaller dataset that is easier to analyse
What are manifest variables?
Your individual questions
What are latent variables?
Your broad factors
What does Kaiser-Meyer-Olkin (KMO) measure?
- Sampling adequacy
* Are there enough to have a reliable solution?
What does a determinant measure?
- Tests for singularity in the data
* Variables should be correlated, but not too much
What does Bartlett’s test of sphericity measure?
- Correlations between clusters of variables
* FA is only appropriate if variables correlate
What is a determinant? (vs sphericity)
- Shouldn’t have all variables correlated.
- If all questions represent the same thing, what is there to factor together?
What is sphericity? (vs determinant)
Must be some correlations between variables.
If each question represents a different thing, how can we make factors?
In a rotation, where do you want the factors to be positioned?
The two factors are far from each other: clearly distinct factors Likely to be a “good” solution
In a rotation, where you you want your items to be located?
Items tightly clustered: Likely to be a “good” factor
What is factor loading?
- Tells you how good the item is within the factor
- Correlation between an individual item and factor it is placed in
- High loadings are good
- Greater than .40 - good enough to be included within a factor
What is split half reliability? (internal consistency)
- Divide items in half: each half should be highly correlated
(consistency of the measure) - Cronbach’s alpha (α): all ways of splitting the items in half combined in one measure – mean split half
