This is a CFA where the model is well specified, as it fits correctly in samples with subjects from several countries. However, in Chile, we find that a latent variable of first order, (In the standardized estimates) presents a negative residual variance and a loading greater than 1. A secondary issue for now, is that the CFI and TLI values are slightly below the acceptable.
I understand that we are facing Heywood cases. Any idea how to solve it? I share with you the MPlus syntax, my database and the model diagram.
Thank you very much.
This issue occurs frequently in higher-order CFA and in bifactor models when they are applied to measurement designs in which the first-order factors represent structurally different constructs or facets. We discussed this issue in detail in
Eid, M., Geiser, C., Koch, T., & Heene, M. (2017). Anomalous results in G-factor models: Explanations and alternatives. Psychological Methods, 22(3), 541-562.
It essentially means that your g2 factor is identical with the first-order factor f5 (f5 has no residual variance = no specificity). For more details, see the above paper.
Hola Flavio, tengo la sensación de que Christian está en lo cierto.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire