15.7 What to watch out for

  • Number of factors should be chosen with care. Check default options.
  • There should be at least two variables with non-zero weights per factor
  • If the factors are to be correlated, try oblique factor analysis
  • Results usually are evaluated by reasonableness to investigator rather than by formal tests
  • Motivate theory, not replace it.
  • Missing data - factors will only be created using available data.