On the replication of factor structures of the Positive and Negative Affect Schedule (PANAS)

- We investigated two-factor and more complex models of the PANAS.
- Measurement invariance of the PANAS was tested for gender and presentation form.
- The PANAS structure is invariant across gender for different factor models.
- Paper-pencil forms should be preferred over online PANAS versions.
- A bifactor model revealed best model fit of the PANAS.

This study focuses on the replication and comparison of different factor structures of the Positive and Negative Affect Schedule (PANAS), a widely used adjective-based questionnaire which assesses positive affect (PA) and negative affect (NA). Four of the previously most prominent proposed factor models were compared by means of confirmatory actor analysis (CFA) for a trait-like instruction in a German sample (n = 413). The results showed poor fit for the two-factor models and an improved fit for the three-factor models. According to RMSEA and conventional cutoff values, only the bifactor model had an acceptable fit. The fact that only a bifactor model fits to the data indicates that the PANAS has structural ambiguity. Moreover, loading invariance as well as partial intercept invariance were demonstrated for all models by means of multiple-group CFA for gender, but it was less clearly found for presentation form (paper-pencil vs. online). This implies that results from paper-pencil presentation of the PANAS probably generalize across gender but do not necessarily generalize to the online presentation form.