CFI/TLI are implausibly high, do not match Stata

I've yet to figure this one out and it's been an issue going back to the 2017 *Socius* paper. Fitting these models in `lavaan` gives CFI/TLI (and probably any related fit index) values that are almost always extremely close to 1. I don't really know how to approach coming up with an alternative approach as I'm relying on `lavaan` to calculate fit indices. I assume there's an issue with how `lavaan` is defining the null model, but I don't know what `lavaan` sees as the null model or what an appropriate null model would be.

Start by reproducing the CFI/TLI results from the R package's dynamic panel model fitting and compare them with Stata. Read how lavaan defines the null model and fit indices, then determine and validate an appropriate alternative; done means the indices are no longer implausibly close to 1 and agree with the reference behavior.

Written by the indexing model from the issue text.