Currently I am working on my master thesis. I have a moderated-mediation model, so used process macro model 7.
However, when I did a check with Two-way Anova to check if the significance from IV-M with MOD is the same, I got different p-values. Both show insignificance for IV-MED (p = 0.941) in two-way ANOVA and (p=0.146) in Process macro, but it worries me that both tests (which must show the same results) does not show the same p-values.
Does anyone knows why this happens, if this is normal and I can ignore it or what I can do to change it?
Thank you in advance!
DISCLAIMER: I do not use the PROCESS macro. However, if model 7 is used for moderated-mediation models, how do you propose to estimate an equivalent moderated-mediation model using "two-way ANOVA" with a native SPSS command (presumably the UNIANOVA command)?
Perhaps you mean that you want to estimate a model without the mediation component using both PROCESS and the UNIANOVA (or REGRESSION?) command in SPSS? If so, you need to show us your command syntax and output plus the output from PROCESS. If you do that, someone may be able to help.
PS- My shot in the dark guess is that interacting variables were centered in one case but not in the other. If so, the first order effects for the interacting variables will be different in the two analyses.
Hello Bruce Weaver ,
Thank you for your answer! I tried to compare results of the moderation part of my model (in my moderation-mediation model) by comparing results of Univariate and process macro, just for checking if it gives the same answer. But you were correct, mean centering was not activated in my process macro. P-values are almost similar now: ANOVA gives (p=0.941) and Process macro gives (p = 0.884). These discrepancies seems okay, right?
If you estimate the same model in the same way (with variables centered the same way, etc.), you should get exactly the same p-values from PROCESS and UNIANOVA. As I said in my first reply, you need to show us the output from PROCESS and you need to show us your syntax and output for UNIANOVA.
Thanks for sharing some output. The R2 and RMSE values for the two models are the same, so you are estimating the same model. But to make the comparison easier, you need to ask UNIANOVA to give you a table of parameter estimates (i.e., coefficients). To do that, add this sub-command: /PRINT=PARAMETER. Then compare the results in that table to the coefficients PROCESS is showing you.
Bruce Weaver Thank you so much for your help and time! This clarifies a lot!!
You're welcome, Simge Akar. For the benefit of others who are following this thread (or who find it later), you might consider posting an image showing the parameter estimates from UNIANOVA so that readers can compare with the estimates from PROCESS. ;-)
- Badges
- Science method