Hi! The weakest items can be identified from the factor loading. if the loading is less that 0.4 or 0.5 considered as low loading(Hair et al., 2010). Thanks

Hello Artem,

In general, nobody pays much attention to the p-value of the overall chi-square test for model-data fit, because it is so sensitive to sample size (and, with even the most minute discrepancy in model-data fit, with some N the result will be significant). RMSR and RMSEA are much better options. If you'd like a quick overview of the families of fit indices, you may find this link to be helpful: http://www.davidakenny.net/cm/fit.htm

As Shamima suggests, individual indicators/items for your latent variables that have the lowest estimated loadings are those that pull down the average variance accounted for, so these might be potential candidates for modification or discarding. However, do be careful as these decisions mean that you are changing your measure(s) based on the data, so the possibility of overfitting (and thereby reducing generalizability) is a very real concern.

Good luck with your work.

Dear Shamima,

Thanks for your comment.

Dear David,

Thank you for your answer as well.

I have read some studies regarding the influence of the sample size on the p-value but I was certain that it is only a case when sample size is large while in my case I have only 146 cases.

What's more, I have already checked the loadings during my CFA and all loadings were significant.