How large was the estimated latent correlation and was it statistically significantly different from zero? What was your sample size? Did the model otherwise fit well and result in plausible parameter estimates and standard errors? Could there be a data error or recoding (item polarity) problem?

Christian Geiser

Thank you for your responding. May I give you the more detail as following

1. I proceeded the short-FFMQ model by using AMOS with 537 samples.

2. Before adjusting the model, nonjudging facet was negative correlated to observing, describing, actawareness, and nonreating (r= -.871, -.770, -.250, & -.862 with p < .01 respectively).

3. Nonjudging components or observed variables including 5 items have standardized factor loading as .43, .64, .41, .35 & .22

4. Model fit indices are not meet the criteria of good fit (relative chi-square = 5.5, GFI = .799, NFI .692, TLI = .693, CFI = .731, RMSEA = .092, RMR .071).

5. Yes, the original items of nonjudging facet are negative/reversed items. However, I correctly recoded them before CFA conducting.

The fit doesn't look good. This could result in biased and unexpected parameter estimates. You should look into the causes of the misfit. I would not interpret the estimated for the current misspecified model.

This is a sign of model misspecification.

Try constraining the correlation between the latent variables using CFA.

Christian Geiser Chuck A Arize

Many thanks for your suggestions.

So that the modification of the model is needed to improve the fit, right?

In this case, can I delete the nonjudging facet from the model if it can produce the good fit.

Christian Geiser

As I mentioned before, I am facing a problem regarding the unexpected negative correlation among components of the Five Facets of Mindfulness Questionnaires in CFA. The model does not meet the good fit indices and has small factor loadings.

The components of FFMQ include observing, describing, actawareness, nonjudging, and nonreacting. However, unexpected correlations were observed: 1) actawareness showed a slight correlation with observing and nonreacting facets; 2) nonjudging exhibited a moderate correlation with observing, describing, nonjudging, and nonreacting, but showed a positive correlation with act awareness. In fact, the five facets of mindfulness should be positively correlated with each other.

Based on this event, I decided to address the issue of careless responders by excluding those who demonstrated insufficient effort. This decision was made due to the scale's inclusion of both positive and negative items. Furthermore, aligning with the theoretical perspective of FFMQ, I also opted to remove subjects who provided extreme responses or rated the items in the actawareness facet in contrast to those in the observing and nonreacting facets. After removing these subjects, the sample size is now 500 cases.

Subsequently, during the preliminary examination, I observed that act awareness is positively correlated with observing, describing, and nonreacting, but it does not exhibit a correlation with nonjudging. Furthermore, the CFA model achieved a good fit, better factor loading, and demonstrated construct validity after deleting the subjects and excluding the nonjudging facet from the model.

Here is my question…

"Is the method I applied for deleting the subjects or responders acceptable in an academic article? Moreover, do you know of any principles that support this method if it is acceptable?" Moreover, do you have any suggestions?

Thank you very much in advance for your suggestion.