Moderated mediation in SPSS using Hayes' Process Macro (Model 7)

[@SM2708-cg4wg]

Dear Dr Crowson. Superb explanation and a brilliant PPT covering all aspects. Thank you

[@mumtazkhan8572]

Thanks, I was greatly worried about learning moderated mediation and your video helped me to learn it. I thank you from the core of my heart.

[@Artyom109Zinchenko]

Hi Mike, thanks a lot for the great video. I just wanted to note that you could get the lines in the initial SPSS plot by clicking 2 times on the image and selecting the corresponding option (It is a small image of a scatterplot with multiple lines). That is, no need to use the extra plotting option from SPSS. Thank you once again.

[@milliondollarman123]

Excellent demonstration!

[@mikecrowson2462]

I'm so glad you found it helpful Qais! Best wishes!

[@shannawielinga8207]

Dear Dr. Crowson, thank you so much for this clear and elaborate instruction on how to perform and interpret moderated mediation. I have a question about including covariates. You can do this immediately in the PROCESS tool, but I wonder what is the best procedure for using the build-in function in the PROCESS tool to determine whether to include them?

[@MANSIGUPTAPHD-]

Hey can you explain modell 11

[@globaljourneywithmancy]

Thank you so much for this video. I was wondering how to interpret the findings if the interaction coefficient value of moderator is in negative.

[@raz2936]

This is so well and clearly explained!!! Thanks!!!

[@bagusadi1277]

Can the results of ordinal data (data obtained from Likert scales) be processed using PROCESS analysis, especially the mediation moderation model in model 7?

[@shamsbharwani4887]

Do you have any videos on PROCESS Model 59? That would be great. I am working on my dissertation, and I am testing the moderation of the mediation effect using Model 59. Thanks

[@mamangefendy]

Thank a lot, verry helpfully

[@iceaapjuh]

Hello Mike, I have a question. I have a conceptual model like this but my W does not moderate path a but B. So I'm using another template, but the idea is the same right? And .. I have not 1 X-variable, I have 3 X'es. They all have an direct effect on Y, and they are all mediated by one M-variable. But how do I run this in SPSS? I can only put in one independent variable? Do I have to add my other 3 X'es as covariates?

[@merveyuksel3048]

Hi Dr. Crowson, can moderator variable in this model be categorical? and could you give an example about that?

[@bipolarluck]

Hello Dr. Crowson, is it possible to have to moderators in the path a? If so, how should I run my analysis? Thank you for the video!

[@napatpaisatchayakul4343]

Hi! Dr. Crowson, Thanks for the valuable video! May I ask you if my X is 1(showing a) and 0 (not showing a) and also if my W or moderator is 1(showing b) and 0 (not showing b) how could I prepare the data before running on this model? Thank you in advance for your help. :) hope to hear from you!

[@MrArdahazal]

Hi, Can I ask a question: In moderated mediation analysis, is it necessary for the direct effect of the mediation (M) on the dependent variable (DV) to be statistically significant, or is It only sufficient for the interaction effect of the moderated variable (W) and mediator (M) on the DV to be significant.

[@user-rmjvkrf]

Thank you, Dr. Crowson, for these valuable videos! Would you consider making a video on Monte Carlo simulations for moderated mediation model power analysis? It would be much appreciated :)

[@nikollouzi8694]

Thank you for your help!!

[@MariaIrimciuc]

Hello Dr. Crowson! Can you help me to elucidate this error: INDIRECT EFFECT: Error encountered in source line # 49752 Thank you!

[@madhumaurya3968]

Thank you sir,nice explanation.. please share something on table structure or reporting data

[@rogerturingan6141]

this is very helpful video sir . May I ask if we still need to conduct preliminary analysis like checking of assumptions before conducting these and do we still need to report model fit indices? Thanks!

[@mikecrowson2462]

Hi Roger. It's always a good idea to check assumptions. And of course, things get muddier with this kind of model since to do it you would essentially have to run separate regression analyses for each submodel if you were seeking to ascertain possible issues with heteroskedasticity, presence of outliers and influential cases, normality of residuals (not quite as problematic with larger n's), linearity, and non-collinearity. [Funny that there's not much out there speaking to the need to do this, especially since the regressions in Process are using OLS regression and would be generated using the standard regression menus in SPSS.] One nice thing with Hayes Process is that it DOES include the option (options menu / heteroskedasticity-consistent inference) for Heteroskedasticity-consistent standard errors to be computed for the OLS regression submodels. I would opt for HC3 or HC4. But basically, to perform a more complete set of diagnostics, you'd need to do it with each regression submodel. I have a few videos you might find useful on regression diagnostics: Assessing heteroskedasticity: kzbin.info/www/bejne/jaCsdHyOg7aNhbM Assessing multicollinearity: kzbin.info/www/bejne/fpnXpWeujrablZI Assessing influential cases and outliers: kzbin.info/www/bejne/q529foqprtmrfqs Cheers!

[@rogerturingan6141]

That is a very informative response... Thanks and looking forward for more great videos... More power to you sir!

[@youngrankim9278]

I wonder difference between SEM and process macro model7.

[@elisabethstrijdhorst]

Dear Dr. Crowson, thank you for this video. I have a question about my output, especially the model summary. There are no 'R' or 'R-sq', but a '-2LL' and 'ModelLL'... what does this mean? The outcome variable is binary and I used model 1 (because I'm only looking for a direct effect and moderation). I hope to hear from you!

[@mikecrowson2462]

Hi Elizabeth. If you have a binary outcome variable and using model 1, then you are effectively doing a moderated binary logistic regression model. There are no traditional R and R-squares for this kind of model. The -2ll is called the model deviance and is not very useful in itself. The fit of the full model is tested using a likelihood ratio chi square test. That's what is reported with the model chi square. I believe Process also provides McFadden's pseudo R-square. It's an analogy only and should not be treated as indicating proportion of variance accounted for. I'd encourage you to check out my video on binary logistic regression here: kzbin.info/www/bejne/rJLFaoGbr92jeMU I hope this helps!

[@elisabethstrijdhorst]

@@mikecrowson2462 Thank you for your clear answer! I'm continuing with the video you recommend!

[@madhumaurya3968]

Sir, the value ur selected..Is this scale value or z value.. please make it clear

[@mikecrowson2462]

Hi there, if you are referring to the conditioning values, then I selected 1sd below the mean, the mean, and 1sd above the mean on the centered moderator. The conditioning values selected will be values on your centered moderator. If it helps, think of it this way, where sd = standard deviation of the centered variable (which is same as the noncentered moderator): mean-1z*sd = mean-sd 0 mean+1z*sd = mean+sd I'm not sure if this is what you were asking, but it's what I believe you were asking :) Cheers!