Confirmatory Factor Analysis (CFA) is a set of multivariate techniques aimed at validating the relations between the observed variables, or indicators, and underlying latent variables, or factors, and is typically used to describe the underlying structure of psychological scales and other social science measurements.

Confirmatory Factor Analysis (CFA) is a set of multivariate techniques aimed at validating the relations between the observed variables, or indicators, and underlying latent variables, or factors. The relations between factors are left unmodeled (and a larger structural equation model, or SEM, could be used to explain these relations, if needed), while the relations between the factors and the indicators are usually assumed linear (although some nonlinear measurement models, such as item-response theory, or IRT, can be formulated as binary item factor analysis models).

CFA is typically used to describe the underlying structure of psychological scales and other social science measurements. The output of the procedure is usually threefold. First, the overall fit test describes how well the assumed (covariance) model fits the data. Second, the individual coefficient estimates may be of interest (e.g., answering the questions like, which variable is the best indicator of a given latent factor). Third, factor scores can be obtained for the elements in the sample based on the observed indicators, to summarize their underlying traits (depression level, IQ, math achievement in psychological and educational contexts; democracy or corruption in sociology and political science; etc.)

Related tags:

  • (structural equation modeling),
  • (encompasses both CFA and its cousin, exploratory factor analysis),
  • (item-response theory)

Synonym: cfa