Exploratory factor analysis
Explore a plausible latent measurement structure
Exploratory factor analysis models shared covariance among indicators using a smaller set of latent factors. Unlike PCA, it distinguishes common from unique variance. Extraction, number of factors and rotation require reasoned choices; oblique rotation allows correlated factors. EFA is exploratory even when software produces precise loadings and attractive factor labels.
Use when several indicators may reflect underlying dimensions but the structure is insufficiently specified for confirmation. Ensure the indicators cover the construct and the sample can support stable covariance estimation.
Strengths
- Reveals patterns of shared indicator variance
- Supports assessment and refinement of measurement structure
Limitations
- Factor retention and rotation involve judgement
- Solutions can be sample-specific and difficult to replicate
Know the boundary
A factor label is an interpretation, and EFA alone does not establish that a psychological entity exists.