Anonymous posted on Sunday, January 25, 2004 - 3:30 pm
We are presently planning a family study where we would like to use multi-level SEM to examine potential influences on the psychosocial adjustment of children (level 1) nested within family systems (level 2). In the current or upcoming versions of Mplus, is there a way to estimate nonrecursive models? We have not seen this issue mentioned in any of the references we have collected on multi-level SEM. Thank you in advance for whatever help you can provide. Sorry, I mistakenly posted this under the discussion on power.
I'm a licensed Mplus user and in my study tested a non-recursive model with:
X -> Y -> Z -> X
I obtained a full circle (cyclic relationships) with path coefficients greater than .60. However, a reviewer commented that cyclic relationships are problematic. Although I made an extensive search, I couldn't find out what he/she means. I know that we typically use recursive models but why one should not use nonrecursive ones?
I'd appreciate if you throw some light to get the point?