It is common for misfit to occur with SEM models. Firstly, it is important to check your model makes theoretical sense and that there are no jarring errors. It is essential to be aware of the nature of your data (i.e., missingness, continuous/categorical data, etc.) to select the most appropriate estimator. Also be aware that SEM typically requires much larger sample sizes compared to other less-complex methods, depending on the number of parameters and model complexity (250+ is very common to see in published articles). Make sure everything is coded and entered correctly before building your SEM model. You could also (cautiously) look at the modification indices your tool produces to see where model improvements can be made - but be mindful of non-generalizability. SEM is a theory-driven approach, so if your model doesn't fit well, it is likely because the framework you identified doesn't match what your data is doing.

If your output is mostly randomness it's more importante to describe the histogram, usually fit models works just for behavours with low randomess.

Hi, Ahmad Aloran ! It all depends on the objective of your research and the theory you're working with. As Ilia Marcev said, if your CB-SEM model is too complex (too may constructs) for your sample size, it will not have a proper fit. You can try to reduce the complexity of your model if the theory allows you to do so. Another solution is to change the SEM method, and apply a PLS-SEM analysis, which requires the analysis of predictive power of the model not the fit of the data. But, in this case, you must change the objective of your research from the confirmation of your theory to the prediction of your endogenous construct.