seminar 2 factor analysis Flashcards
What is factor analysis?
A statistical method which looks at how lots of different items correlate and determines how many theoretical constructs could most simply explain what you see.
What is principle component analysis?
Very similar to factor analysis, we conduct PCA rather than FA as it is less complex and is psychometrically sound
What are the differences between Factor Analysis and Principle component analysis?
FA uses a mathematical model from which the factors are estimated
PCA uses the original data to derive the set of clusters of variables
What does factor analysis do?
- Helps us determine items/behaviours relating to constructs
- Takes info and simplifies it by placing it into factors
- Examines the pattern/
correlations between variables to calculate new variables (super variables or FACTORS) - Determining the maximum amount of common variance using the smallest amount of explanatory constructs
What are the uses of FA?
- Understanding the structure of an underlying dimension/ construct
- to construct a questionnaire
- to reduce a data set to a more manageable and purposeful size.
How can factor analysis help with questionnaires
Can determine which behaviours add up to a personality trait.
What can we infer by looking at the correlation between scale items?
Something about their underlying nature (theoretical constructs)
What causes increase chance of a type 1 error?
More correlations- so we don’t just use a correlation matrix
How do you conduct factor analysis?
- input the data into spss
- conduct FA (PCA)
- check assumptions
- How many factors? (Eigenvalues and scree plot)
- Re-analyse with rotation
- Interpret factors
- factor scores
What is the correct way to input data in spss for fa?
A score should be supplied for each question, opposed to an overall score
How do you conduct FA in spss?
- > Analyze
- > dimension reduction
- > factor
- > Select descriptives, univariate and first column of correlation matrix
- > select extraction and then scree plot
- > options: sorted by size and suppress small coefficients, change value to .40
What is the Barlett’s test?
Measures whether the correlation matrix differs from any identity matrix and therefore, should be significant
What is the KMO test?
Indicates if there is a distinct and reliable set of factors from the patterns of correlations between variable- this will lie between 0-1, look for a value closer to 1.
What is the eigenvalue?
The variance accounted for by that factor. Commonly, only factors of eigenvalues above 1 should be considered.
What is Kaiser criterion?
Eigenvalues
Why should kaiser be used with caution?
Criterion is accurate with less than 30 variables and communalities >.70
What does a scree plot do?
Lists factors in order of eigenvalues.
What is the cut off point of a scree plot?
The point of inflexion (where the point of the line changes/ flattens)
Why are un-rotated factors difficult to interpret?
1- factors correlate with many variables
2- variables load onto many factors
What does rotation do?
Maximises loading of each variable on to one of the extracted factors and minimises loadings on the other
What are the different types of rotation?
Orthogonal
Oblique
What is orthogonal rotation?
We do not assume out factors to correlate. Varimax, Quartimax, Equamax
Which type of orthogonal rotation did we use and why?
Varimax which minimises the humber of high loadings on a factor and maximises the difference between dimensions.
What is oblique rotation?
We assume our factors to correlate (good theoretical evidence suggesting so). Oblimin, Promax.
