Thank you for your helpful and clear video. Could you please explain how you know the indirect effect of X on Y is significant? Is it simply because the effect of .4278 falls between the confidence interval lower and upper limits of .1311 and .8349 respectively? Thank you for you help.
@mike Crowson in the second model with 2 mediators: how do we interpret the direct effect? I notice it is no longer sig. Total mediation? Is it because of the second mediator? I'm confused because the second indirect path was not significant. Perhaps it still contributes to the model? Huge fan, btw. Please don't stop.
Hello. Thanks for this video. How to check relationship between mediator and binary outcome (when mediator is continuous) and X and Y relationship when X is continuous? Is η correlation correct?
Dear Professor, could you tell me how to interpret the results mediation (using PROCESS) for a binary IV? I have a negative direct and indirect effect, and I wish to know to which level of the IV (manipulated in an experiment) can this negative value be attributed..in other words - do both of the IV conditions have a negative effect on the DV? or is it one of them? Thank you !
4:32 Is "direct effect" the correct terminology to be used. As path 'a' builds upon the "indirect effect" whilst the term "direct effect" relates to the main path 'c'' IV to DV?
This is great--there aren't many resources for how to interpret PROCESS analyses with a binary outcome. Thank you! I was wondering: are there any issues with multiplying the coefficients from a linear OLS regression and a logistic regression when calculating the indirect effects? If so, how should one deal with that?
What if all the three components X, M, and Y are binary? Is that considered a violation when using the bootstrapping method? Thanks for your clarifications
Dear Mike, I guess I'm wrong... Why is the direct effect from IV epe to the DV is named political interest? I hoped it would be donate? The path (direct effect) from external political efficacy to political interest is positive and significant (b=.4439, s.e.=.2220, p=.0456)??
Hi there. There's a typo in there. The second part of the sentence is correct, but the reference to the dv in the first part is incorrect. Sorry for your confusion on this. Cheers!
@@mikecrowson2462 Thank you so much for your quick reply. So many people watch your videos 👍. So it helped me checking over and over before asking. It is so great that you can use the ppt and copy the reporting analysis phrases. All the best for 2021!!
Dear Professor, could you tell me how to interpret the results mediation (using PROCESS) for a binary IV? The IV is abuse no=0 and yes=1 (2 categories) on AV Group 1/2 with Mediation IV Attachment (continuous). The coeff of abuse on group is -.800, what does that say abuse no 0 leads to being in group? I am really confused
Thank you Mike for your awesome videos ! I have a question. When running a mediation analysis using macro process, how are we supposed to calculate the r-square of the dependent variable? It seems macro process doesn't calculate it.
Hi Armin, I'm glad you have found them helpful! Regarding your question: You will obtain an R-square for the mediating variable (since it relies on OLS regression); just look at the part of the output containing the F-test results. If your DV had been treated as continuous, then the R-square would also appear along with F-test results because OLS regression would have been performed. In your case with the binary outcome, Process registers the DV as binary and then uses logistic regression on the second part of the model. The only R-square values you will find are 'pseudo-R-squares' (that's the McFadden, Cox & Snell, and Nagelkerke values you will see). These are analogies to the OLS R-square and are not directly comparable since they aren't computed the same way. However, they are typically used descriptively as another piece of information when evaluating the fit of a logistic regression model. If you want to know more related to logistic regression, check out the following video: ruclips.net/video/cpWSSJHuT2s/видео.html Also, I talk a bit more about the logistic model in the context of moderated medation models using Process in a new video here: ruclips.net/video/FPCTXcxQFgg/видео.html Best wishes!
Hi Mike, thank you for this video. May I ask if you know how to derive log odds statistics that are more relevant for a binary outcome? Thank you once again :)
Hi Jomel, are you asking about how to obtain the log odds (Y=1) for each case with respect to the outcome variable? If so, you would need to run the second model as shown in the Hayes output (where the dv is the binary variable) through the binary logistic (under the regression) menu in SPSS. When you are specifying the model and options, click Save and select probabilities (under predicted values). These are the predicted probabilities(Y=1) for the cases. You can obtain the predicted logits by using the Transform->Compute function in SPSS where you create a new variable (you might call it 'logit') and specify the expression as: ln(pre_1/(1-pre_1)) where pre_1 is the predicted probability generated using your logistic regression. There you go. You now will have the predicted logits (Y=1) for each case. Just keep in mind that this really has nothing to do with the mediation analysis and is only predicted logits coming from that second logistic regression model. I hope this helps!
Hi Mike, Thank you, very helpful. I was wondering, in my output with PROCESS v3.3 version, I've got an output with the standardised b, yet do I use the unstandardised b or the standardized b? Also because my mediator (categorical yes/no) is measured differently than the IV (5-point Likert scale). Other research papers used the previous version of PROCESS, but that version didn't show the standardised b, that's why I'm confused. Also, my direct effect is not significant (IE= -.0723, p = .678 )and (95% CI= -.4147, .2701), yet my indirect effect is significant (IE= .124) and (95% CI= .0260, .2626). The total effect model isn't displayed. So, do I have partial mediation? You didn't explain the direct effect in the video or powerpoint. Much appreciated, Lianne Engwerda (University of London)
Hi Celine, I don't believe you are able to have the mediator or Y variable be multinomial using Process. I know you can use Process where the independent variable is categorical and if you want to test a moderator hypothesis with a categorical moderator. But I think your categorical Y variable can only be binary. If I find out anything different I will try to post something on that. Best wishes.
Is there a way to run mediation in SPSS with a binary mediator or would you recommend another statistical package, such as Mplus? Alternatively, could I run a linear model testing the a path first and then the b path as a starting point? If the a or b path are not significant, then I would not expect mediation to occur, correct?
dear dr. Crowson, thank you so much. I was wondering how iI get a p value for my indirect effect? I know it is significant as my the confidence interval is greater than 0.00 , but I would like to have a p value as well. Thanks, Sanne
Hi Sanne, thanks for your question. No p-value is printed out because you are not using a theoretical sampling distribution where the p-value would be computed from. The bootstrap procedure used in Process relies on a percentile bootstrap method, where the confidence interval is formed by taking the estimated indirect effect from the empirical sampling distribution that falls at the 2.5th percentile and the 97.5th percentile of that distribution. In some programs such as AMOS and Stata (both of which also use - or at least allow for - the percentile bootstrap method), it is possible to also obtain a p-value. However, that approach relies on a bootstrapped estimate of the standard error of the indirect effect. That bootstrapped estimate is used in conjunction with the normal distribution where the null of ie=0 falls at the center and the bootstrapped standard error can be used to generate a z-value, from which the p-value is derived. This is a normal theory method of testing the indirect effect. To tell the truth, I'm not a big proponent of using this approach. The sampling distribution of the indirect effect tends to be asymmetric since the indirect effect is computed as the product of paths a and b (see MacKinnon, 2008), which would seemingly make a normal theory method (which assumes symmetry of the sampling distribution, where symmetry is a component of normality) problematic. This problem with asymmetry is also a major reason why we have moved away from using the Sobel test (using the delta method for computing standard errors) to relying on procedures such as the percentile bootstrap. Anyway, this is probably way more than you wanted to know. However, it was fun thinking through this - especially since I am primed on this after re-reading MacKinnon (2008). Cheers!
@@mikecrowson2462 Dear dr. Crowson, thank you very much for you kind and clear answer. To be aware of the above mentioned theory as background knowledge will improve my statistical choices. I will stick to the CI :) Best regards, Sanne
Thank you so much for this helpful video for my Masters thesis! I was just wondering how I should interpret the output in SPSS when my independent variable is multicategorical? Because there are 2 different interactions showing up in my output and I don't know what those mean... Greetings from Belgium!
Hi Marie, if you use the Categorical / Indicator option in Process for your IV, then it will recode your IV into dummy variables (where the # of dummy variables will equal the # of groups minus 1; it also appears that the group on the original IV with the lowest value will be treated as the reference category). You can think of the interaction terms as breaking down the total interaction between the IV and moderator into separate components, with each potentially predicting your outcome variable. I don't have any videos or demonstrations exactly addressing your question. However, I do have a video demo on moderated logistic regression by checking this video (ruclips.net/video/Dbwq_or9I0M/видео.html), where the moderator is actually treated as continuous. Also, if you are testing interactions in the context of mediation, then you are basically specifying moderated mediation. I have a video on this topic in the context of a binary outcome here: ruclips.net/video/vW69r24hPs4/видео.html . I hope this is helpful to you!
Very good video, thank you very much ! However I have a question : when you insert 2 mediators in the process 3.5 how do you compare their indirect effects to evaluate their relative importance ? thank you in advance for your answers
Hi there. Thanks for your question. I'm assuming you are using process model #4 with parallel mediators. Believe it or not it's pretty easy to test the difference in indirect effects. Under Options click on "pairwise contrasts of indirect effects.' When you run the analysis, you will get the indirect effect via each mediator. There will be a contrast estimate provided with a bootstrap confidence interval of the estimate. As an example of what you might see, below is some output I generated for a two mediator model. C1 is the difference between the two indirect effects. The bootstrap confidence interval does not contain the null of 0. So the two indirect effects are significantly different from each other. Indirect effect(s) of X on Y: Effect BootSE BootLLCI BootULCI TOTAL .0635 .0130 .0391 .0907 engage .0510 .0109 .0313 .0753 interest .0125 .0079 -.0027 .0278 (C1) .0385 .0139 .0125 .0677 Specific indirect effect contrast definition(s): (C1) engage minus interest I hope this helps. Cheers!
With this method, would you be able to determine the R-squared (variance explained) of individual predictors? I notice it gives you the r-squared of the whole model, but what about for each individual predictor? Or would I need to run a separate multiple regression to find that? (I am doing this for a study I'd like to publish, thanks in advance!)
Hi Andrew, it gives you the pseudo R-squares for each regression in the model (keep in mind these are rough analogies to R-square, as they are not computed the same way as OLS regression; I discuss that here: ruclips.net/video/cpWSSJHuT2s/видео.html). If your DV was continuous, then you'd be able to re-run each regression model in your analysis and request the semi-partial correlation between each IV and the DV - and squaring this would produce the increment in R-square uniquely accounted for by a given predictor. Unfortunately, there is no corresponding approach with logistic regression. If you aim is to determine the relative contributions of predictors, you could always standardize the IV's before entry, and you'd end up with a partially standardized solution (see e.g., ruclips.net/video/P0VKXrHrFUc/видео.html). I hope you find this helpful. Best wishes!
sir, what do we do if the dependent variable is measured on a categorical scale ( Yes, No, Not calculated), and the rest independent variable and mediator are measured on an interval scale? please sir, waiting for your reply.
I don't really understand what the categories on the dv represent so it makes answering tricky. If your DV is binary, you can still use process to test for mediation. The second regression equation with the binary outcome will be for binary logistic regression. If your categorical DV contains more than 2 levels, then Process has no direct mechanism for testing mediation in this case. There is no multinomial or ordinal logistic regression capabilities using Process
Thankyou sir for replying🙏😊 Actually my DV is about having sufficient fund for post-retirement life among employees. Sir can we make this model in PLS-SEM? Actually I couldn't find any video on RUclips about this.
hello sir, your all video is very helpful for me. i am using model no 6 with 2 mediators. but face problem in the interpretation of data. if possible kindly share related video link.
Dear Mike, thanks for the videos. Do you know how to use the same sample each time in mediation? I read something about the seed function but I do not know how to use it
Many thanks. Very helpful.
You are welcome, Snow! Thanks for visiting!
The best most simple explanation of mediation I've ever seen,
Big help for my diss, cheers
The most useful interpretation about mediation that I have ever watched, thanks for helping!
Extremely helpful! Thank you for this.
Thanks for visiting, Pamela! best wishes!
Thank you very much sir 🙏
Thank you and it is really helpful!
thank you sir. very helpful
Very helpful, thank you :)
Thank you Professor. Content was accurate and easy to follow and implement. Great work.
This is great! Thank you.
Thank you for your helpful and clear video. Could you please explain how you know the indirect effect of X on Y is significant? Is it simply because the effect of .4278 falls between the confidence interval lower and upper limits of .1311 and .8349 respectively? Thank you for you help.
@mike Crowson in the second model with 2 mediators: how do we interpret the direct effect? I notice it is no longer sig. Total mediation? Is it because of the second mediator? I'm confused because the second indirect path was not significant. Perhaps it still contributes to the model? Huge fan, btw. Please don't stop.
Hello. Thanks for this video. How to check relationship between mediator and binary outcome (when mediator is continuous) and X and Y relationship when X is continuous? Is η correlation correct?
Professor, thank you so much for this video! it helped me with my Master's thesis a lot!
Glad it helped Zarina! Best wishes to you!
Much appreciated!
You're very welcome!
Dear Professor, could you tell me how to interpret the results mediation (using PROCESS) for a binary IV? I have a negative direct and indirect effect, and I wish to know to which level of the IV (manipulated in an experiment) can this negative value be attributed..in other words - do both of the IV conditions have a negative effect on the DV? or is it one of them?
Thank you !
4:32 Is "direct effect" the correct terminology to be used. As path 'a' builds upon the "indirect effect" whilst the term "direct effect" relates to the main path 'c'' IV to DV?
This is great--there aren't many resources for how to interpret PROCESS analyses with a binary outcome. Thank you! I was wondering: are there any issues with multiplying the coefficients from a linear OLS regression and a logistic regression when calculating the indirect effects? If so, how should one deal with that?
What if all the three components X, M, and Y are binary? Is that considered a violation when using the bootstrapping method? Thanks for your clarifications
Dear Mike,
I guess I'm wrong...
Why is the direct effect from IV epe to the DV is named political interest? I hoped it would be donate?
The path (direct effect) from external political efficacy to political interest is positive and significant (b=.4439, s.e.=.2220, p=.0456)??
Hi there. There's a typo in there. The second part of the sentence is correct, but the reference to the dv in the first part is incorrect. Sorry for your confusion on this. Cheers!
@@mikecrowson2462 Thank you so much for your quick reply. So many people watch your videos 👍. So it helped me checking over and over before asking. It is so great that you can use the ppt and copy the reporting analysis phrases. All the best for 2021!!
Dear Professor, could you tell me how to interpret the results mediation (using PROCESS) for a binary IV? The IV is abuse no=0 and yes=1 (2 categories) on AV Group 1/2 with Mediation IV Attachment (continuous). The coeff of abuse on group is -.800, what does that say abuse no 0 leads to being in group? I am really confused
Thank you Mike for your awesome videos !
I have a question. When running a mediation analysis using macro process, how are we supposed to calculate the r-square of the dependent variable? It seems macro process doesn't calculate it.
Hi Armin, I'm glad you have found them helpful! Regarding your question: You will obtain an R-square for the mediating variable (since it relies on OLS regression); just look at the part of the output containing the F-test results. If your DV had been treated as continuous, then the R-square would also appear along with F-test results because OLS regression would have been performed. In your case with the binary outcome, Process registers the DV as binary and then uses logistic regression on the second part of the model. The only R-square values you will find are 'pseudo-R-squares' (that's the McFadden, Cox & Snell, and Nagelkerke values you will see). These are analogies to the OLS R-square and are not directly comparable since they aren't computed the same way. However, they are typically used descriptively as another piece of information when evaluating the fit of a logistic regression model. If you want to know more related to logistic regression, check out the following video: ruclips.net/video/cpWSSJHuT2s/видео.html
Also, I talk a bit more about the logistic model in the context of moderated medation models using Process in a new video here: ruclips.net/video/FPCTXcxQFgg/видео.html
Best wishes!
@@mikecrowson2462 Thank you for such an informative response.
@@mikecrowson2462 Hi Mike, is it better to perform hierarchical linear regression or to perform simple linear regression for each variable separately?
Thank you so much sir you explained very simply,i was helped me a lot
Hi Mike, thank you for this video. May I ask if you know how to derive log odds statistics that are more relevant for a binary outcome? Thank you once again :)
Hi Jomel, are you asking about how to obtain the log odds (Y=1) for each case with respect to the outcome variable? If so, you would need to run the second model as shown in the Hayes output (where the dv is the binary variable) through the binary logistic (under the regression) menu in SPSS. When you are specifying the model and options, click Save and select probabilities (under predicted values). These are the predicted probabilities(Y=1) for the cases. You can obtain the predicted logits by using the Transform->Compute function in SPSS where you create a new variable (you might call it 'logit') and specify the expression as: ln(pre_1/(1-pre_1)) where pre_1 is the predicted probability generated using your logistic regression. There you go. You now will have the predicted logits (Y=1) for each case. Just keep in mind that this really has nothing to do with the mediation analysis and is only predicted logits coming from that second logistic regression model.
I hope this helps!
Dear Mike, when a coefficient is greater than 1, what does it indicates?
Hi, is it now possible to run mediation with a categorical mediator variable using SPSS process?
Hi Mike,
Thank you, very helpful.
I was wondering, in my output with PROCESS v3.3 version, I've got an output with the standardised b, yet do I use the unstandardised b or the standardized b? Also because my mediator (categorical yes/no) is measured differently than the IV (5-point Likert scale).
Other research papers used the previous version of PROCESS, but that version didn't show the standardised b, that's why I'm confused.
Also, my direct effect is not significant (IE= -.0723, p = .678 )and (95% CI= -.4147, .2701), yet my indirect effect is significant (IE= .124) and (95% CI= .0260, .2626). The total effect model isn't displayed. So, do I have partial mediation? You didn't explain the direct effect in the video or powerpoint.
Much appreciated,
Lianne Engwerda (University of London)
Dear professor, it is possible to test mediating effect using Andrew Hayes Process in multinomial logistic regression? Thank you very much.
Hi Celine, I don't believe you are able to have the mediator or Y variable be multinomial using Process. I know you can use Process where the independent variable is categorical and if you want to test a moderator hypothesis with a categorical moderator. But I think your categorical Y variable can only be binary. If I find out anything different I will try to post something on that. Best wishes.
Is there a way to run mediation in SPSS with a binary mediator or would you recommend another statistical package, such as Mplus? Alternatively, could I run a linear model testing the a path first and then the b path as a starting point? If the a or b path are not significant, then I would not expect mediation to occur, correct?
dear dr. Crowson, thank you so much. I was wondering how iI get a p value for my indirect effect? I know it is significant as my the confidence interval is greater than 0.00 , but I would like to have a p value as well. Thanks, Sanne
Hi Sanne, thanks for your question. No p-value is printed out because you are not using a theoretical sampling distribution where the p-value would be computed from. The bootstrap procedure used in Process relies on a percentile bootstrap method, where the confidence interval is formed by taking the estimated indirect effect from the empirical sampling distribution that falls at the 2.5th percentile and the 97.5th percentile of that distribution. In some programs such as AMOS and Stata (both of which also use - or at least allow for - the percentile bootstrap method), it is possible to also obtain a p-value. However, that approach relies on a bootstrapped estimate of the standard error of the indirect effect. That bootstrapped estimate is used in conjunction with the normal distribution where the null of ie=0 falls at the center and the bootstrapped standard error can be used to generate a z-value, from which the p-value is derived. This is a normal theory method of testing the indirect effect. To tell the truth, I'm not a big proponent of using this approach. The sampling distribution of the indirect effect tends to be asymmetric since the indirect effect is computed as the product of paths a and b (see MacKinnon, 2008), which would seemingly make a normal theory method (which assumes symmetry of the sampling distribution, where symmetry is a component of normality) problematic. This problem with asymmetry is also a major reason why we have moved away from using the Sobel test (using the delta method for computing standard errors) to relying on procedures such as the percentile bootstrap. Anyway, this is probably way more than you wanted to know. However, it was fun thinking through this - especially since I am primed on this after re-reading MacKinnon (2008). Cheers!
@@mikecrowson2462 Dear dr. Crowson, thank you very much for you kind and clear answer. To be aware of the above mentioned theory as background knowledge will improve my statistical choices. I will stick to the CI :) Best regards, Sanne
Thank you so so much, this was extremely helpful!! I could use the help for my thesis. :)
is there a possibility to include control variables using this method? such as age, education etc.?
Thank you so much for this helpful video for my Masters thesis! I was just wondering how I should interpret the output in SPSS when my independent variable is multicategorical? Because there are 2 different interactions showing up in my output and I don't know what those mean... Greetings from Belgium!
Hi Marie, if you use the Categorical / Indicator option in Process for your IV, then it will recode your IV into dummy variables (where the # of dummy variables will equal the # of groups minus 1; it also appears that the group on the original IV with the lowest value will be treated as the reference category). You can think of the interaction terms as breaking down the total interaction between the IV and moderator into separate components, with each potentially predicting your outcome variable. I don't have any videos or demonstrations exactly addressing your question. However, I do have a video demo on moderated logistic regression by checking this video (ruclips.net/video/Dbwq_or9I0M/видео.html), where the moderator is actually treated as continuous. Also, if you are testing interactions in the context of mediation, then you are basically specifying moderated mediation. I have a video on this topic in the context of a binary outcome here: ruclips.net/video/vW69r24hPs4/видео.html . I hope this is helpful to you!
@@mikecrowson2462 legend - this is the key to my conundrum. All your videos hit the nail on the head! Big up - your a number one G
Very good video, thank you very much ! However I have a question : when you insert 2 mediators in the process 3.5 how do you compare their indirect effects to evaluate their relative importance ? thank you in advance for your answers
Hi there. Thanks for your question. I'm assuming you are using process model #4 with parallel mediators. Believe it or not it's pretty easy to test the difference in indirect effects. Under Options click on "pairwise contrasts of indirect effects.' When you run the analysis, you will get the indirect effect via each mediator. There will be a contrast estimate provided with a bootstrap confidence interval of the estimate. As an example of what you might see, below is some output I generated for a two mediator model. C1 is the difference between the two indirect effects. The bootstrap confidence interval does not contain the null of 0. So the two indirect effects are significantly different from each other.
Indirect effect(s) of X on Y:
Effect BootSE BootLLCI BootULCI
TOTAL .0635 .0130 .0391 .0907
engage .0510 .0109 .0313 .0753
interest .0125 .0079 -.0027 .0278
(C1) .0385 .0139 .0125 .0677
Specific indirect effect contrast definition(s):
(C1) engage minus interest
I hope this helps. Cheers!
With this method, would you be able to determine the R-squared (variance explained) of individual predictors? I notice it gives you the r-squared of the whole model, but what about for each individual predictor? Or would I need to run a separate multiple regression to find that?
(I am doing this for a study I'd like to publish, thanks in advance!)
Hi Andrew, it gives you the pseudo R-squares for each regression in the model (keep in mind these are rough analogies to R-square, as they are not computed the same way as OLS regression; I discuss that here: ruclips.net/video/cpWSSJHuT2s/видео.html). If your DV was continuous, then you'd be able to re-run each regression model in your analysis and request the semi-partial correlation between each IV and the DV - and squaring this would produce the increment in R-square uniquely accounted for by a given predictor. Unfortunately, there is no corresponding approach with logistic regression. If you aim is to determine the relative contributions of predictors, you could always standardize the IV's before entry, and you'd end up with a partially standardized solution (see e.g., ruclips.net/video/P0VKXrHrFUc/видео.html). I hope you find this helpful. Best wishes!
sir, what do we do if the dependent variable is measured on a categorical scale ( Yes, No, Not calculated), and the rest independent variable and mediator are measured on an interval scale? please sir, waiting for your reply.
I don't really understand what the categories on the dv represent so it makes answering tricky. If your DV is binary, you can still use process to test for mediation. The second regression equation with the binary outcome will be for binary logistic regression. If your categorical DV contains more than 2 levels, then Process has no direct mechanism for testing mediation in this case. There is no multinomial or ordinal logistic regression capabilities using Process
Thankyou sir for replying🙏😊
Actually my DV is about having sufficient fund for post-retirement life among employees.
Sir can we make this model in PLS-SEM?
Actually I couldn't find any video on RUclips about this.
Hey Mike thanks for the video. What are some assumptions that must be met in order to perform this analyses?
These*
Dear MIke, is it possible to carry out a mediation analysis involving a binary outcome variable with process version 4.0?
Hi Raj, as far as I am aware you should be able to. Cheers!
@@mikecrowson2462 Thank you very much for clarifying doubt. This was extremely helpful 🙏
hello sir, your all video is very helpful for me. i am using model no 6 with 2 mediators. but face problem in the interpretation of data. if possible kindly share related video link.
Dear Mike, thanks for the videos. Do you know how to use the same sample each time in mediation? I read something about the seed function but I do not know how to use it
I ran the PROCESS syntax file from the macros zipfile download, but how do you get it to show up as an option under "regression" tab in SPSS?
Hi there. I have a video on install the macro to use under the regression tab here: ruclips.net/video/11tNWOJPCzo/видео.html
@@mikecrowson2462 Awesome! Got it- thank you!
How can I interpret a coefficient when under this situation (with binary DV) it becomes more than 1?
Hello, is it also possible to use Process when you have an dichotomous X and Y? It doesn't work by me..
same here
Please help in moderation mediation and mediation moderation. Thank You
Hi Professor, please advise whether it's possible to use Process with IV binary -> M continuous -> DV binary? thanks in advance!
Hi Sofia. Sure that should be fine. Cheers!