|
|
 |
| Indirect effects with binary mediator... |
 
|
| Message/Author |
|
|
| George Howe posted on Wednesday, June 03, 2020 - 8:56 am
|
|
|
I ran a moderated mediation model with continuous cause and moderator, binary mediator and outcome. The model parameters are significant for moderation of the a path and for the simple b path, suggesting significant moderated mediation.
MODEL INDIRECT statement was for effect estimates across the moderator ranging from -1 to +1 SD. The first table gives estimates of indirect effect parameter at that level of the moderator; all of the indirect effects across all levels are nonsignificant. For example:
Two-Tailed Rate of
Estimate S.E. Est./S.E. P-Value Missing
Effects from PB to SISCAT for BHS = -1.000
Tot natural IE 0.059 0.050 1.173 0.241 0.570
Odds ratios for binary Y
Tot natural IE 2.071 0.543 3.812 0.000 0.527
Does the second table provide log odds ratios for the same direct and indirect effects?
If so, those are are all significant for the indirect effect across all levels of the moderator. Any thoughts as to why the indirect effect parameters are all nonsignificant while the log odds ratios are all significant? |
|
|
|
|
| Not sure what you mean by the first and second table. Effects are given in probability metric and also in odds ratio metric; is that what you mean? I don't know if you use bootstrapping which might be needed for probability effects given non-symmetry of CIs. We need to see the output pieces you point to. |
|
|
| George Howe posted on Thursday, June 04, 2020 - 12:31 pm
|
|
|
Bengt,
How do I send the output? I tried to include it in my post, but it is way more than the 1.5K allowed.
On bootstrapping: these are based on imputed data, so I get a message that MPLUS can't do bootstrapping under those circumstances. Is there a way to do that?
Thanks,
G |
|
|
|
|
Send to Support@statmodel.com
Let me think about the second question. |
|
|
| Back to top |
|
|
