 Asseming the feet of structural ecosystem models requires some expertise. Let's take a look at summary of how you should be assessing these models. My recommendations follow client's recommendations. I think he argues his case really well in the book and he presents these five points. The first point is that you should respect the chi-square and report the chi-square value, its significance and its decrease of freedom. If the chi-square is significant, then that should be acknowledged in the paper and used to do diagnostics. Particularly in large samples, when chi-square fails the model, then you need to look at the individual or residual covariances and assess whether the differences, whether they are small enough that they can be considered trivial or not. If your sample size is small and chi-square does not reject the model, the diagnostics are nevertheless required because it is possible that the model is mis-specified and the mis-specification would be revealed by the residual covariances, but you just don't have enough power in the chi-square test to detect those mis-specifications. So in any case, chi-square reporting and diagnostics is the point number one. Then, if the model does not fit the data well or there is some decurred mis-specification, use the report, the residual covarious matrix or at least describe what part of the model, the data, what part of the data the model does not explain. And this allows your readers to make and have an informed opinion on whether you have done a good job in detecting the possible mis-specifications and also if the mis-specifications can influence your results. If you choose to report on the alternative or approximate fitting disease, then you should go with this set. But this is generally not there for diagnostics purposes, but it is more as a matter of convince. Quite often when you do testing the model test, chi-square rejects the model. Then you look at the residuals and you realize that there is something that should have been in the model that you didn't put there in the first place. So you have a theoretical reason to re-specify the model. All these theoretical reasons for model re-specification, if you add something, should be reported along with the rationale for doing so. If you choose to retain a model where the chi-square is significant so there is nothing reasonable that you can do, then you need to provide evidence for example in the form of residual covariances that the decree of misfit is really small and you can do for example sensitivity analysis by freeing some of the parameters to see how that influences the results. Finally, if your model is conclusively rejected by the chi-square statistic, it does not mean that your research is necessary, completely untrustworthy. That your model is rejected, not supported by the data, is also a finding itself and then you can start interpreting what does the rejection of your model mean for the theory being tested.