Mediation analysis - Hayes Process plugin

Question

I have a model to test 4 hypothesis (all data was collected on a 5-point likert scale through surveys):

1. Ability (IV) has a positive relationship with PO fit (M)

2. PO fit (M) has a negative relationship with uncertainty (DV).

3. Ability (IV) has an inverse relationship with uncertainty (DV)

4. PO fit (M) mediates relationship between Ability (IV) and uncertainty (DV).

I'm using the Hayes PROCESS macro on SPSS to run the mediation analysis, which gives me the following 5 results:

A. The effect of Ability (IV) on PO fit (M) (path a)

B. The effect of PO fit (M) on uncertainty (DV) while controlling for Ability (IV) (path b)

C. The effect of Ability (IV) on uncertainty (DV) while controlling for PO fit (M), also called the direct effect (path c')

D. The total effect of Ability (IV) on uncertainty (DV) (path c)

E. The indirect effect of Ability (IV) and PO fit (M) on uncertainty (DV) (path a * path b)

Now, the Hayes plugin can help me conclude on hypothesis 1 through result A, hypothesis 3 through result D, and hypothesis 4 through result A-E. But how do I conclude on hypothesis 2 on PO fit (M) has a negative relationship with uncertainty (DV)?

The Hayes plugin will only let me get result B, which is The effect of PO fit (M) on uncertainty (DV) while controlling for Ability (IV). But I need the effect of PO fit (M) on uncertainty without controlling for Ability (IV) ie., without Ability being present.

Is there a setting in Hayes that lets me get to this result? If not, must I run a separate linear regression just for the effect of PO fit (M) on uncertainty without controlling for Ability (IV)?

Answer 1

I'm not sure why you would want this relationship, as it is clearly confounded by the very assumption that your mediation model is plausible. You can estimate the relationship between M and DV in a separate regression (that is all PROCESS is doing anyway), but that association tells you little when you know part of that association is due to the IV (i.e., because the IV is a common cause of M and DV). If the PROCESS macro does not give you that value, it is with good reason. It is not a value commonly reported in mediation analysis as it has little useful interpretation. The b path is more informative. Ideally every path is estimated adjusting for confounders; otherwise your entire mediation analysis is useless anyway.