Process Module
Good morning! I have an issue with the process module that is confounding me. I'm running a series of simple models: One IV, one DV, one Mediator, and one Moderator (3 levels, nominal). Requesting Hayes output 15 or 59. In my output, the direct, indirect, and total effects for the first level of my moderator are identical to those for the third level of my moderator--always, across five different IV's and two different mediating variables. I know that this can be a sign of no mediating effect, but this doesn't match up with output from Mplus, and it seems unlikely to be occurring across all these analyses. And by identical, I mean the estimate, standard error, z-value, p value, and lower/upper 95% CI's.
Any thoughts on why this might be happening?
Comments
Difficult to say. Can you share the data or show a screenshot?
There are two separate issues: which paths are moderated, and how the moderator is coded.
In PROCESS model 15, W moderates the M → Y path and the direct X → Y path. In model 59, W also moderates the X → M path. For a nominal moderator with three levels, W should not be entered as a single numeric variable unless the codes represent a real ordered, equally spaced trend. A three-category nominal moderator requires two contrasts; coefficients and conditional effects depend on the contrast system and reference category.
Before comparing JASP with Mplus, align the model number, reference category, contrast coding, nominal/continuous treatment, centering, covariates, missing-data sample and standardized versus unstandardized output. Conditional effects at the mean and ±1 SD are meaningful for a continuous moderator, not an unordered nominal moderator.
If point estimates still differ, write out both equations and the exact contrast matrix. Bootstrap choices may change confidence intervals, but should not produce different point estimates for the same fitted model.