Faktorscores

Faktorscores, in English often called factor scores, are numerical estimates of latent factors produced by factor analysis and related latent-variable models. They represent estimated values of underlying constructs for each observation, derived from observed variables that load on the same factor. Factor scores enable researchers to use latent constructs in downstream analyses (for example, regression or clustering) without re-estimating the factor model for every analysis.

Estimation of factor scores involves computing linear combinations of observed variables with weights derived from the factor-loading matrix and the chosen extraction/rotation method. Common approaches include regression-based scores, Bartlett’s method, and other score-estimation procedures. Because different estimation methods yield different scores, factor scores are not uniquely defined for a given solution.

Factor scores are estimates, not direct measurements. They depend on the assumed factor model, the number of factors extracted, and the rotation employed. They are scaled and centered according to the method used, and their interpretation refers to the latent construct rather than individual observed items.

Factor scores have several limitations. They incorporate measurement error from the observed variables and may be biased if the model is misspecified or loadings are unstable. When comparing scores across groups, researchers should check model fit and measurement invariance and clearly report the estimation method used.

In practice, factor scores are used in psychometrics, education research, and social sciences to summarize latent constructs such as ability, motivation, or attitudes and to serve as predictors or outcomes in further analyses.