Actually this is quite common in a CFA processing. Here the model fit only presents the information about if the theoretical model is acceptable, which means it only concerns about if the given model is the best one, or to say, an acceptable one, but not including if the factor loading is good enough. It should be noticed that there's not a certain correlation between the model fitness and the factor loading. While the theoretical model could be the best solution, but the actual measurements of some items may not be that accurate to the model, the situation you mentioned would occur. In some papers, the Extroversion from the FFM was reported the same situation as you observed.

It is hard to say the scales you used is not valid as well. But for most cases, you should process an EFA to verify if the low-loading items are not suitable for the measurement in your sample. If so, you could just delete these unacceptable items then do the CFA again, and in the final paper you need to report the information about the revised version and how you revised it. It is noticeable that this series of operation could limit the academic value more or less, which could affect the range of journals you submit the study. If the result of EFA is not the same as the CFA, probably you need to check if you did the CFA incorrectly. 

Hope my answer is useful for you!

Good luck~

Factor loads less than 0.50 are not so suitable, have you done EFA before? Is your theoretical model strong enough for the factor structure you are in?

What type of variable you have?

What estimator used?

What method of rotation?

[email protected]

Hug

Actually this is quite common in a CFA processing. Here the model fit only presents the information about if the theoretical model is acceptable, which means it only concerns about if the given model is the best one, or to say, an acceptable one, but not including if the factor loading is good enough. It should be noticed that there's not a certain correlation between the model fitness and the factor loading. While the theoretical model could be the best solution, but the actual measurements of some items may not be that accurate to the model, the situation you mentioned would occur. In some papers, the Extroversion from the FFM was reported the same situation as you observed.

It is hard to say the scales you used is not valid as well. But for most cases, you should process an EFA to verify if the low-loading items are not suitable for the measurement in your sample. If so, you could just delete these unacceptable items then do the CFA again, and in the final paper you need to report the information about the revised version and how you revised it. It is noticeable that this series of operation could limit the academic value more or less, which could affect the range of journals you submit the study. If the result of EFA is not the same as the CFA, probably you need to check if you did the CFA incorrectly. 

Hope my answer is useful for you!

Good luck~

I have the same problems. But when I run the CFA by single latent, it would be very low loading. Then I did EFA, the dimension split to two dimension and finally I develop two laten construct. Then the loading factor increased but the problem is data is not significant after that. If I compute in SPSS, I think that no need to split the item because it is just reverse item. Not a different themes.

Any item having a factor loading less than 0.6 and an R2 less than 0.4 should be deleted from the measurement model.

Assalam alykom Usama Rehman

actually bro now am facing same problem, if you solved it, plz help me.