Mediation, moderation, and measuremen...
Message/Author
 Student posted on Thursday, February 26, 2015 - 8:19 am
Hi Drs. Muthen,
I am relatively new to Mplus and wondered if I am performing/thinking about these analyses correctly. I have a mediational model (X->M->Y) with a covariate. I tested the mediational model with the MODEL INDIRECT.
My first two questions are: I have 1) only put the covariate on M and Y (i.e., M on covariate, Y on covariate) and not X, given previous Mplus threads, and 2) also used BOOTSTRAP methods; is this what you would recommend?
Next I was interested in if there are any differences across gender and performed MULTIPLE GROUPS (with "GROUPING IS gender (1=male 2=female);" but no bootstrap). My model is indicated as:
Y by y1 y2 y3;
X by x1 x2 x3;
M by m1 m2;
cov; !to handle missing data on x
Y on M (ymmale) cov;
M on X (xmmale) cov;
MODEL female:
Y on M (ymfem) cov;
M on X (xmfem) cov;
MODEL CONSTRAINT:
NEW (ind_1 ind_2 diff);
ind_1 = ymmale*xmmale;
ind_2 = ymfem*xmfem;
diff = ind_1-ind_2;
 Student posted on Thursday, February 26, 2015 - 8:19 am
Continued from above:
I understand that this method allows you to see if there are differences across gender in the indirect effects. My third question is, is bootstrapping recommended here?
Lastly my fourth general question is: I am stuck on whether it is necessary to determine whether there is invariance across gender and how to do so. I understand needing to test for invariance in the 1) general factor structure, 2) factor loadings, 3) intercept, then 4) residual variance. I did test for measurement variance initially this way using the MODEL INDIRECT, but my results (e.g., for the indirect and direct effects) are completely different compared to the results I get when I used the code above using MODEL CONSTRAINT.
Any help would be greatly appreciated on how to proceed-- thank you.
 Bengt O. Muthen posted on Thursday, February 26, 2015 - 10:16 am
We request posting in only one window.

1) Right

2) Yes

You cannot give more than one label before;
so this type of statement doesn't do the job:

Y on M (ymmale) cov;

As for your 4th question you want to ask this type of general analysis strategy question on SEMNET.
 Student posted on Thursday, February 26, 2015 - 11:09 am
Thank you for your response and apologies for posting on more than one window.

I wondered if you could expand on what you mean by "You cannot give more than one label before; so this type of statement doesn't do the job:"
Is there some other way you recommend to test for gender differences in mediation?

Thank you.
 Bengt O. Muthen posted on Thursday, February 26, 2015 - 11:42 am

Y on M (ymmale) cov;

you should say

Y on M (ymmale)
cov;

So you split the statement into two lines with only one label per line. This is necessary unless separation by semicolon (;) like

y (yvar); m (mvar);
 Student posted on Thursday, February 26, 2015 - 12:29 pm
Thank you!