I have a construct measured by 28 items on Likert scale. Principal component analysis with promax rotation produced 4 factors with two of them having just two items on each. After eliminating those items and one item that had loading over one as well as one item that didn't load over any factor, the factor structure is now composed of two factors with 19 items loading on the first factor and three items loading on the second one. Is it ok to have too much items loading on one factor compared to the other, or should I do the analysis again and enforce one factor?
I would think about how the three items that load on the second component might differ substantively from the other items before "forcing" all of the items onto a single component. Also, loadings > 1 are possible and admissible with oblique rotation (correlated components). So dropping the item with the loading > 1 may not be necessary.
Thanks Geiser. All the items that didn't load onto the main factors (including those that were eliminated except the one with loading greater than one) all of them is reverse items. Interestingly, this applies to all other constructs in the instrument, reversed items are all liading on one factor. Testing hypotheses keeping the two factors solution produces a standardized path coefficient greater than one
It is common for reverse coded items to load on a separate factor. Confirmatory factor analysis models are better suited to address this issue, for example by including additional method (residual/specific) factors for one set of items (e.g., the negatively worded items). See, for example:
Geiser, C., Eid, M., & Nussbeck, F. W. (2008). On the meaning of the latent variables in the CT-C( M -1) model: A comment on Maydeu-Olivares and Coffman (2006). Psychological Methods, 13(1), 49–57.
Herzberg, P. Y., Glaesmer, H., & Hoyer, J. (2006). Separating optimism and pessimism: A robust psychometric analysis of the revised Life Orientation Test (LOT–R). Psychological Assessment, 18, 433–438.
Excuse me Dr. Geiser, Would you provide me with a reference saying that factor loadings can exceed one with oblique rotation. I need to cite this.
- Badges
- Science topic
- Similar topics
- Filing