Confirmatory factor analysis (CFA)

Exploratory factor analysis looks for groups of questions in the data. CFA works the other way round: you specify which questions belong to which factor, for example from a theory or an existing questionnaire, and test how well that fits the data. The result is the model fit (fit indices) and the standardized loadings.

Example: Nine questions, three factors of three questions each: CFI = .93, RMSEA = .09. The fit does not meet the rules of thumb: CFI below .95, RMSEA clearly above .06.

Caution: A good fit does not show that the factors measure the right thing, and another model can fit equally well. If the model is adjusted after looking at the same data, the test is no longer independent.