I am using the modification indices with the WLSMV estimator (dichotomous variables). I ran analyses sequentially freeing parameters (item loadings) to find a Partial Scalar Invariant model. I completed the analysis twice once using the M.I. value in the modification indices and once using the StdYX E.P.C. The two approaches produced varying results. Is there any research or does the Mplus team know which modification indice value is preferred as the basis of selecting the parameters to amend in subsequent analyses?