STATS 10 EXPLORATORY FACTOR ANALYSIS Flashcards
What are the uses of EFA?
EFA is useful for finding a multiple-factor model that best explains the pattern of correlations between observed variables
What are factorially simple variables?
Variables loaded with only one factor e.g. most psychological measures
What are factorially complex variables?
Variables loaded with more than one factor e.g. a problem in a maths test that involves both algebra and geometry
What are the residuals in factor analysis?
Residuals in factor analysis are differences between the observed correlations and correlations predicted by the model
How do residuals help determine the number of factors?
Large residuals in a 1-factor model show that we need another factor, you then repeat the residuals check for the 2-factor model and accept it if they’re good
What is the rotation problem?
In factor models, the axes can be altered in a way that keeps the data points constant, so that we end up with a more desirable rotation.
What is rotation?
Rotation is a transformation to achieve factorially simple variables from a multiple factor model (usually)
Which rotation method should you use?
An oblique rotation (e.g. oblimin, promax) which allows axes to change from their 90 degree angle
Why are factor correlations always slightly higher than correlations between sum scores?
because factor correlations are between latent factors that have taken an estimate of error into account
When would we use the maximum likelihood estimation method?
For large samples and normal data
When would we use the unweighted least squares estimation method?
For small samples and non-normal data
What is measurement by modelling?
Factor analysis, provides a scaling method
What sample size do you need for factor analysis?
AT LEAST 200, unless you have really “strong” data