Mher B. posted on Saturday, November 30, 2013 - 4:34 pm
Dear professors, Accidentally,I have noticed an intriguing problem. When I am changing the sequence of the equations for measurement model (BY statements) in the model command I am getting different results in a structural part (ON/BY statements): a non-significant path coefficient (endogenous ON b*exogenous) becomes significant or vice versa. The model contains ESEM component but the issue manifests itself not only when changing the order of equations (BY statements) of the same exploratory set but also shifting equations not in an exploratory set. Can you think of possible explanations, in general, what can be the source for this instability in Mplus 7? In general, what steps one should do in this case? (e.g. a simplistic circumvention - change the sequences, stick with the most conservative solution if there are no other problems with model specification and estimation).