Using Mplus You can do it via MODEL CONSTRAINS: Using Karin's notation You should declare new variables:

a1b1=a1*b1;

a2b2=a2*b2;

ab=a1b1 - a2b2;

In output file there would be additional parameters a1b1,  a2b2 and ab. If parameter ab will be significantly different from 0 the indirect effects differ.

maybe this link will be helpfull

Available from: https://www.researchgate.net/post/How_can_I_test_whether_two_indirect_effects_are_significantly_different_in_SEM [accessed Mar 23, 2015].

best regards 

magdalena

Stephanie,

yes, you can do that. One possibility would be to constrain the path coefficients pertaining to the indirect effect to be equal across both groups.

If you have, for example, a path from predictor X to mediatior M in group 1, a1: X -> M, and from mediator M to outcome Y, b1: M-> Y, and the same paths in group 2, a2 and b2, you could include a constraint such as a1 = a2 and b1 = b2. Comparing the chi-square values (chi-square difference test) of the model with freely estimated coefficients and the model with the constraints would provide a chi-square value of the difference. With 2df the difference should be larger than 5.991 (see chi-square table).

Does this help?

Karin

I don't see any requirement for you to make that comparison. Yes you can try using the method in Amos called "Manage Model". You constrain one and evaluate the other and vice versa.

Using Mplus You can do it via MODEL CONSTRAINS: Using Karin's notation You should declare new variables:

a1b1=a1*b1;

a2b2=a2*b2;

ab=a1b1 - a2b2;

In output file there would be additional parameters a1b1,  a2b2 and ab. If parameter ab will be significantly different from 0 the indirect effects differ.

In support of Grzegorz Humenny's answer, I LOVE using Mplus for any SEM problem - it is worth the learning curve to use it.

Process provides information on pair-wise indirect effect comparisons in the output. See Hayes 2018 book.