unsupervised
Exploratory factor analysis
Estimate latent factors that explain correlations among observed numeric variables.
Requirements
- several related numeric variables
- adequate sample size
- factorable correlation matrix
Avoid when
- variables are unrelated
- a purely predictive reduction is the goal
Formula
x = Λf + ε
Interpretation
Loadings describe each variable's association with the retained latent factors.
Worked example
Estimate latent dimensions underlying responses to a multi-item questionnaire.
Open example datasets