I conducted a confirmatory factor analysis (CFA) using the Adverse Childhood Experiences International Questionnaire (ACE-IQ) to assess its underlying structure. I specifically examined a three-factor model comprising Household Dysfunction, Abuse, and Neglect as latent factors. The resulting CFA model plot, which is attached below, revealed some negative factor loadings and factor correlations that warrant clarification.
I would greatly appreciate insights from the experts in this forum to help me understand the implications of these negative factor loadings and correlations present in the model plot. I am seeking constructive feedback and guidance on whether it is appropriate to report these results as they are or if there are adjustments or interpretations that should be considered.
Thank you in advance for your valuable input and expertise in this matter.
Item 3 and 4 are not negative statements.
It is difficult to say without knowing what the items are. The negative loadings are small (near zero). This indicates that those items are not really measures of the neglect factor. You should base the interpretation of the factors on those items that have decent (large) loadings.
With regard to the factor correlations, are you concerned about the negative correlations between the neglect factor and the abuse and dysfunction factors? One question would be how well the model fits in terms of the chi-square test of model fit. I suspect the fit is mediocre at best given the large number of items. A suboptimal fit can lead to parameter bias, although it seems unlikely that that would entirely explain unexpected negative factor correlations.
Have you tried an exploratory factor analysis (EFA)? EFA might allow you to better filter out those items that are good indicators of the factors versus poor ones and learn more about the factor structure of this scale which may be more complex than the simple structure implied by your CFA.
I agree with @Christian Geiser about the advisability of an EFA. Theoretical preconceptions aren't always correct and correspondingly won't always give a good fit in CFA. However, I would precede that with a parallel analysis to ensure that your EFA reflects the corret number of factors.
Respected Dr. Geiser and Kohn,
Based on your suggestion I ran EFA to explore the underlying factor structure of the ACE-IQ using the r code: efa_result
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India