Gang and I looked at this and considered the computation of the parsimonious fit index. We went back to Bollen's 1986 paper where it says minimally restrictive model, but it's not clear if he meant null model anyway. The Bullmore paper that we based the program on used a null model clearly because a minimum model did not exist. It was only with Stein's paper later that the idea of a minimum model was introduced. Their matlab code uses the null model it seems. In any case, it seems clear that if we did want to continue with the minimum model we would have to scale it properly by the adjusted number of parameters instead of the method currently in the code. We opted to go with your choice and used the chi-square 0 based on the null model. Give the latest version a try, and let us know if it works for you. Check the next binaries post-dating this posting.

Thanks for looking at this and for letting us know.