I am running a path analysis and encountered an issue with the the TLI.
When I run my hypothesized model, removing certain paths (constraining them to 0) seems to lead to an increase in the TLI, which I am finding unusual.
To assess where the problem is, I ran the full saturated model and both my CFI and TLI were 1.00 as expected. However, once I iteratively removed the paths in the model, I noticed that if I remove one specific path rather than decreasing, my CFI remains at 1.00 and my TLI increases to 1.04. It is my understanding that I should see a decrease in both when I remove a path.