Hello Dr. Buchanan, thank you for this video. My X is whether participants have received professional coaching or not, meaning it's categorical but when I choose multicategorical and then indicator I receive and error because it says I need at least 3 categories, would you assist me. What am I doing incorrectly?
This is one of the best explanations I've gotten about how to test for mediation! And I've suffered through 10 graduate level statistics courses. Thank you so much! :) Can you provide access to the Word document? It'd save people tons of time (I especially appreciate the APA format!)
The link provided in the description above takes you to where you can download all the materials provided in these videos along with some example write ups. statstools.com/learn/advanced-statistics/
a step back question, when defining hypothesis do we focus on each relationship, in other words H1; X->M1 / H2: X-> M2 / H3: M1->Y and so on and so forth OR we define them as to H1: M1 mediates the r-ship between X and Y ; H2: M2 mediates the r-ship between X and Y;. For me personally, the first way makes more sense especially after seeing this analysis, but it yet confuses me quite a bit...
Not totally sure I understand the question but the first way combined with the second way makes sense too - as M1 mediating the relationship may require the first hypotheses to be accurate?
That really depends on your hypotheses for how all these variables relate. I would only use this analysis if it answers the research questions you collected that data for.
What if part of our conceptual model is like model 6, but we also have another mediator separately below it which goes from X to Y (model 4), so we have 3 mediators, 2 of them are successieve and 1 separate, how should I analyse it?
Thx for your help , i have a question if a run the model and only change M2 , ¿it is possible that path B1 got a different result or it will always have the same result?
Do you think that it is possible to have video regarding to the Model 14 with PROCESS, in fact I am struggling with the interpretation of the output. Thanks for your videos.
Hi Erin, Thank you so much for this video (and all the resources you make publicly available - it is incredibly informative and helpful. I have a quick question about power/sample size determination for serial/double mediation using PROCESS. I was wondering, based on this video example, what your effect size F2 was and how many tested and total predictors you put in to calculate the sample? Thanks!
You would use 3 X variables (since the biggest model would have X M1 M2) ... I would estimate the effect size based on previous research (use the eta / r2 option if you don't know the f2 translation).
Hi, thanks for this great video. I have one question to ask. How would you interpret the results if X -> m1 -> y is significant (per confidence interval), X -> m2 ->y is insignificant, and X -> m1 -> m2 -> y is significant? Interestingly enough, If I test the serial mediation in a different order (e.g., X -> m2 -> m1 -> y), then X -> m1 -> y becomes insignificant.
Thank you very much for the video! I have a question: If I have 4 mediators but I am interested only in the following paths - x-m1-m4-y; x-m2-m4-y; x-m3-m4-y - would it be valid to run three separate models? Or I must put everything in one model?
With these specific types of questions, it would almost be better to put them into a structural equation model framework. You'd want to run these separately, but I'm unsure how you'd calculate the indirect effect and it's confidence interval without using the larger SEM framework.
Thank you very much for your answer! So, what you mean is to do the analysis outside PROCESS if I do not want to lose power with all 15 paths implied by model 6 with 4 mediators?
@@darkdistinctplaces Right, I'm saying I'm not sure how you would specify such a model in Process, you'd need to do it as a SEM is likely the best way.
Hi Erin, thanks so much for producing these mediation videos. It's really helping me out with my data analysis for my thesis project. One quick question: you mentioned towards the end of the video something regarding the order of M1 and M2 and possibly trying the analysis again in a different order as in this case it showed no relationship. I'm planning to try a mediation analysis with 3 potential mediators. The mediators will be types of coping styles. Should I be concerned about what order I assign each coping style as M1, M2, and M3? I hope the question makes sense. Thanks again!
Right - I think if there's a good theory about M1 --> M2 --> M3 then use them in that order. Otherwise, you might consider why you'd have a strong argument why they are in the order they are in (since mediation is often a correlational design).
I have a total sample size of 62. 1. How do you check GPower for double mediation and 2. Should I use a different method like autoregressive regression?
Power would be calculated by using the number of predictors (three) and sample size. You would change the options to posterior power in Gpower but the rest of the same options that you seen in the video.
Hi - thanks for the great video! I have some questions (1) does the double mediation model makes sense if I do not have a significant direct effect vom X on y ? Because I am currently having a model and the way x-> m1 -> m2-> y is significant - but no other combination has any significance ... and (2) the strange thing is that I analysed the combination of x->m2->y before in a model 4 analysis were this path was significant but now in the model 6 analysis the same x has no significant effect on m2 and this confused me ...
Different variables in the equation (ie model 4 to 6) will definitely show different results (also, maybe missing values were dropped so it's different sample sizes?). I don't think you have to have a significant X to Y, but it should show at least a big change from X to Y without M and X to Y with M.
Hi Erin, thanks a lot for all your videos it is so helpful! you are talking here about the possibility of a suppression. Under what circumstances is it possible that such a suppression occurs and what are the consequences? Do you know good literature about it? Thanks!
Mediation really is a test of suppression, wherein the variables are so highly correlated they start to cancel each other out. Some great comments and links can be found here: stats.stackexchange.com/questions/73869/suppression-effect-in-regression-definition-and-visual-explanation-depiction
Hey Erin I'm busy writing up my psychology honours thesis, and you've saved my results section! I did have one question that not even my supervisor could answer. I ran a series of serial mediation with two mediators, the only thing that changed was X. However, the relationship between M1 and M2, M1 and Y, and M2 to Y changed slightly. You would think that the predictive value of M1 and M2 would not change. So why do the coefficients vary between models, even though the only thing I've changed is X?
hello Dr. Erin. I have some questions about model 4(parallel mediation). If one indirect effect is significant and the other is not while Total indirect is also significant, should the mediation be concluded as significant? or we should give results separately for each mediation? thanks in adavance for your reply!
I would report each individual mediation effect and the total effect separately, so the reader can decide what they think is important given the results of your study.
Hi Erin, nice lecture. I Have a question about interpretation. Must the total indirect effect of X on Y be significant to be allowed to look further into the specific indirect effects? In my data, total indirect effect CI crosses zero, but some specific indirect effects CIs do not. Hope you have time to help. Thanks for the video anyway!
I would think you could just describe it and see what happens reviewer wise. If the hypothesis predicted the different effects might be different, then it would be confirming that? Hard to know without knowing all the details.
If our conceptual model is such that A has impact on B and C, and then B and C has an impact on D and then D has impact on E, then which option of mediation we have to choose.
Hello, I hope you've resolved your issue. If anything the proper mediation is model 6 with 3 mediators (see Template.pdf). By analyzing the results you can confirm if the described effects are existent and if there're other effects in the interaction.
Never saw this message sorry! I would think you could do it, but I'm not sure how to do the code properly to get the indirect effect - I only use lavaan in R now (do not have access to Amos).
Thank you for your very informative demonstrations. I am interested in an extension of model 6; however, with a moderator (Gender) between X and M1. Is such a model possible? Thanks!
Thanks. I am looking for multiple mediation (i.e., model 6) with a moderator between X and M1 only. In this case, previous research has shown that gender is a moderator for M1 but not for M2.
Generally, people check the bootstrapped confidence interval. If it contains zero, you would say no mediation occurred, while if it does not contain zero, mediation has occurred.
This video is incredibly helpful, thank you! Do you have a sample write-up of how you would report these results in APA format available? Additionally, do you have any resources on how to interpret Model 6 when one's outcome variable is categorical?
There are some examples on the OSF page, but I don't think I have one for double mediation specifically. If the DV is categorical, you should be using logistic regression (as linear regression assumes a scale DV) rather than process.
Statistics of DOOM Thanks! When I ran the analysis with process the output provided was that of a logistic regression - is that going to be an erroneous result? If so, any resources you know of for performing serial mediation with a categorical DV without using process? Thanks again!
Maybe that's been added to the newest process (I am unsure but Hayes website would say) - it wouldn't surprise me if it had been added. Otherwise, I would suggest just running the same regressions yourself one step at a time using the logistic regression options (that would also allow you to check the output you have). The problem would be the bootstrapping mediation effect - and I am unsure a good way to do that that doesn't involve using the paid bootstrapping plug in or R.
Depends on if you think you have two separate mediators (M1 does not predict M2 but X->M1->Y and X->M2->Y) or you think you have serial mediation X -> M1 -> M2 -> Y. I would think that would be based on theory, which way you would expect the relationships to pan out.
Hello again Dr. Erin. According to theory M1 -> M2 must be related. Already a bivariate relationship .68*. But while serial mediation M1 and M2 are not significantly related, X -> M1 -> M2 -> Y indirect relationship is significant according to confidence intervals. Can this result be interpreted as a serial mediator? The total effect and the total indirect effect are significant, while the direct effect is based on meaningless confidence intervals. Thank you very much for your support
Hi professor this was so clear to me to make my master thesis analysis, in this part of my thesis your videos was a coadviser to me, i have a basic question that ı couldnt get a clear answer i made a serial mediation (model 6) like that and their order has a theoretical order, they both significant mediators but when i writing the results i couldn't get a proper APA format (i checked the stat tools ) could you please help me
This really very informative tutorial. i have a question, can u plz specify on which conditions we can say its a serial mediation or double mediation?? all three indirect effects (Ind1, Ind2, Ind3) should be significant?? in my model i got Ind: X---->M1----->Y insignifcant. Ind2: X--->M1--->M2-->Y sig Ind:3 X-->M2-->Y sig kindly help me to consider it a full double mediation while direct effect of X on Y is also insignifcant.
That would depend on the index of mediation provided in the output - sounds like double mediation with the sig paths, but hard to know with out the confidence interval on the indirect effects.
@@pnaracet1562 I would just describe it in your paper as you have here. Sounds like M1 mediation is happening, M2 mediation is happening but not M1 --> M2 since that's usually ind3. So, two single mediations.
Hello. Thank you for your videos, they are brilliant! I am using PROCESS for the first time. When I run two separate mediation analysis (X-->M1-->Y and X-->M2-->Y) both mediators are significant. However, when I run parallel or serial mediation, my mediators are not significant. Can anyone please help me understand why this may be?
Thank you for your response!! This is certainly the case. M1 and M2 are very highly correlated. Please could you tell me what the appropriate way to deal with suppression would be?
If M1 and M2 are that highly correlated, then that implies that they are essentially the same variable (like so interrelated, measuring one of them is enough). I'd either pick one or do two separate mediations and show how they give you the same results.
Excuse me Dr. Erin. I have sent you an email. Could you please check your inbox out and see my research model with regression report and help me understand it? Thanks in advance!
Great video! On which theoretical literature did your tests for outliers (Mahalanobis), normality, linearity, homoscedasticity base? Is there literature I can cite for my master thesis?
thank you for saving my dissertation!!!!!!!
Glad to be of help.
Hello Dr. Buchanan, thank you for this video. My X is whether participants have received professional coaching or not, meaning it's categorical but when I choose multicategorical and then indicator I receive and error because it says I need at least 3 categories, would you assist me. What am I doing incorrectly?
See my other comment! It's not needed to check that box since you only have two.
Hi, do you think it is safe to use two mediators, when M1 is positively correlated with IV & DV, M2 is negatively correlated with IV & DV
I wouldn't think that would matter, just influences the signs for the paths.
@@StatisticsofDOOM thank you, in fact I wrote this to Andrew Hayes never expected a return but he said the same
Great video, thank you very much. Will help with my Honours Thesis results writing.
Excellent!
This was incredibly clear and helpful. Thank you!
This is one of the best explanations I've gotten about how to test for mediation! And I've suffered through 10 graduate level statistics courses. Thank you so much! :)
Can you provide access to the Word document? It'd save people tons of time (I especially appreciate the APA format!)
The link provided in the description above takes you to where you can download all the materials provided in these videos along with some example write ups.
statstools.com/learn/advanced-statistics/
Hi Erin, should we check the homoscedasticity only by putting all the variables together or should we do it for all individually? Thanks
Start with the them all together - if it's a bad chart, you can then try them one at a time to see which one might be the problem.
This is so informative and clearly explained. Thanks!
Thank you!
I'm not getting the option under Probit "PROCESS by Andres F. Hayes model-6". Please suggest get this package into the SPSS software.
That might be dependent on the version you are using - I didn't write the plug in so you would need to contact Hayes for those type questions.
a step back question, when defining hypothesis do we focus on each relationship, in other words H1; X->M1 / H2: X-> M2 / H3: M1->Y and so on and so forth OR we define them as to H1: M1 mediates the r-ship between X and Y ; H2: M2 mediates the r-ship between X and Y;. For me personally, the first way makes more sense especially after seeing this analysis, but it yet confuses me quite a bit...
Not totally sure I understand the question but the first way combined with the second way makes sense too - as M1 mediating the relationship may require the first hypotheses to be accurate?
Dear Erin, the instruction is super useful, I thank you so much and btw your voice sounds good~
I have Eight IVs, Two Mediator, and Four different outcomes (DVs). Does this method could analyze it?
That really depends on your hypotheses for how all these variables relate. I would only use this analysis if it answers the research questions you collected that data for.
What if part of our conceptual model is like model 6, but we also have another mediator separately below it which goes from X to Y (model 4), so we have 3 mediators, 2 of them are successieve and 1 separate, how should I analyse it?
not 100% sure on that one - don't know that it's possible with process.
Thx for your help , i have a question if a run the model and only change M2 , ¿it is possible that path B1 got a different result or it will always have the same result?
Normally, I would say they shouldn't change, but it may be because of the other variables in the model?
Do you think that it is possible to have video regarding to the Model 14 with PROCESS, in fact I am struggling with the interpretation of the output. Thanks for your videos.
Hi Erin, like to ask how to control for covariates in this model? (say, if i am 3 covariates)?
You would put them in the covariates box!
Can anyone tell me the difference between serial mediation and double mediation? Or are they the same?
Same thing!
can't thank you enough!! was so helpful
This was so very helpul! Could you possible describe how to report effect sizes for the overall model? Thanks!
Like the R2 values for the total effect model? Or the indirect effect for the double/serial mediation?
Thank god i found this video
Glad to be of help!
Can this model include a moderator as well along with 2 mediators ?
Not using model 6 in process - you will need to find the model picture that matches which conceptual diagram you are imagining.
Hi Erin,
Thank you so much for this video (and all the resources you make publicly available - it is incredibly informative and helpful. I have a quick question about power/sample size determination for serial/double mediation using PROCESS. I was wondering, based on this video example, what your effect size F2 was and how many tested and total predictors you put in to calculate the sample?
Thanks!
You would use 3 X variables (since the biggest model would have X M1 M2) ... I would estimate the effect size based on previous research (use the eta / r2 option if you don't know the f2 translation).
Hi, thanks for this great video. I have one question to ask. How would you interpret the results if X -> m1 -> y is significant (per confidence interval), X -> m2 ->y is insignificant, and X -> m1 -> m2 -> y is significant? Interestingly enough, If I test the serial mediation in a different order (e.g., X -> m2 -> m1 -> y), then X -> m1 -> y becomes insignificant.
I would just spell it out like you've said it. The serial mediation does appear to be in effect but only one effect pathway works individually.
@@StatisticsofDOOM Then is there any way to test which order is more convincing, for instance, based on the model fit aside from the theory?
@@eskim21 I think theory is a lot better on that idea.
Thank you very much for the video! I have a question: If I have 4 mediators but I am interested only in the following paths - x-m1-m4-y; x-m2-m4-y; x-m3-m4-y - would it be valid to run three separate models? Or I must put everything in one model?
With these specific types of questions, it would almost be better to put them into a structural equation model framework. You'd want to run these separately, but I'm unsure how you'd calculate the indirect effect and it's confidence interval without using the larger SEM framework.
Thank you very much for your answer! So, what you mean is to do the analysis outside PROCESS if I do not want to lose power with all 15 paths implied by model 6 with 4 mediators?
@@darkdistinctplaces Right, I'm saying I'm not sure how you would specify such a model in Process, you'd need to do it as a SEM is likely the best way.
Hi Erin, thanks so much for producing these mediation videos. It's really helping me out with my data analysis for my thesis project. One quick question: you mentioned towards the end of the video something regarding the order of M1 and M2 and possibly trying the analysis again in a different order as in this case it showed no relationship. I'm planning to try a mediation analysis with 3 potential mediators. The mediators will be types of coping styles. Should I be concerned about what order I assign each coping style as M1, M2, and M3? I hope the question makes sense. Thanks again!
Right - I think if there's a good theory about M1 --> M2 --> M3 then use them in that order. Otherwise, you might consider why you'd have a strong argument why they are in the order they are in (since mediation is often a correlational design).
Thanks for the reply!
I have a total sample size of 62. 1. How do you check GPower for double mediation and 2. Should I use a different method like autoregressive regression?
Power would be calculated by using the number of predictors (three) and sample size. You would change the options to posterior power in Gpower but the rest of the same options that you seen in the video.
Hi - thanks for the great video! I have some questions (1) does the double mediation model makes sense if I do not have a significant direct effect vom X on y ? Because I am currently having a model and the way x-> m1 -> m2-> y is significant - but no other combination has any significance ... and (2) the strange thing is that I analysed the combination of x->m2->y before in a model 4 analysis were this path was significant but now in the model 6 analysis the same x has no significant effect on m2 and this confused me ...
Different variables in the equation (ie model 4 to 6) will definitely show different results (also, maybe missing values were dropped so it's different sample sizes?). I don't think you have to have a significant X to Y, but it should show at least a big change from X to Y without M and X to Y with M.
Hi Erin, thanks a lot for all your videos it is so helpful! you are talking here about the possibility of a suppression. Under what circumstances is it possible that such a suppression occurs and what are the consequences? Do you know good literature about it? Thanks!
Mediation really is a test of suppression, wherein the variables are so highly correlated they start to cancel each other out. Some great comments and links can be found here: stats.stackexchange.com/questions/73869/suppression-effect-in-regression-definition-and-visual-explanation-depiction
Hey Erin I'm busy writing up my psychology honours thesis, and you've saved my results section! I did have one question that not even my supervisor could answer. I ran a series of serial mediation with two mediators, the only thing that changed was X. However, the relationship between M1 and M2, M1 and Y, and M2 to Y changed slightly. You would think that the predictive value of M1 and M2 would not change. So why do the coefficients vary between models, even though the only thing I've changed is X?
Likely because of the correlation between X1 and Ms and X2 and Ms? That's my best guess, because it is a bit odd!
hello Dr. Erin. I have some questions about model 4(parallel mediation). If one indirect effect is significant and the other is not while Total indirect is also significant, should the mediation be concluded as significant? or we should give results separately for each mediation? thanks in adavance for your reply!
I would report each individual mediation effect and the total effect separately, so the reader can decide what they think is important given the results of your study.
excuse me, please. could not you check your email? I have sent you my results. I would really appreciate if you can give some feedback!
Hi Erin, nice lecture. I Have a question about interpretation. Must the total indirect effect of X on Y be significant to be allowed to look further into the specific indirect effects? In my data, total indirect effect CI crosses zero, but some specific indirect effects CIs do not.
Hope you have time to help. Thanks for the video anyway!
I would think you could just describe it and see what happens reviewer wise. If the hypothesis predicted the different effects might be different, then it would be confirming that? Hard to know without knowing all the details.
If our conceptual model is such that A has impact on B and C, and then B and C has an impact on D and then D has impact on E, then which option of mediation we have to choose.
I'm not sure - I would look at the template.pdf and see which picture best matches what you are trying to describe.
Hello, I hope you've resolved your issue. If anything the proper mediation is model 6 with 3 mediators (see Template.pdf). By analyzing the results you can confirm if the described effects are existent and if there're other effects in the interaction.
Thanks for the video. Do you have information on how to report the Indirect Effects of X on Y?
I usually report them as "the indirect effect was XX 95% CI [ XX, XX ]" filling in the XX with the indirect and the bootstrapped confidence interval.
Hi Erin, just wondering if you can help introduce how to produce a table and illustrate in article based on model 6 result?
Like a table of the results? I imagine you could put in b, SE, t, p for each path and call it good? (not sure I'm following what you are asking).
Hi Erin
I am wondering if double mediations can be carried out using Amos? Do you have a related video on this? Thanks!
Never saw this message sorry! I would think you could do it, but I'm not sure how to do the code properly to get the indirect effect - I only use lavaan in R now (do not have access to Amos).
Thank you for your very informative demonstrations.
I am interested in an extension of model 6; however, with a moderator (Gender) between X and M1. Is such a model possible? Thanks!
Yes, if you wanted to do the steps manually - looks like process version 2 only has the serial mediation but not serial mediation + moderation.
Thanks. I am looking for multiple mediation (i.e., model 6) with a moderator between X and M1 only. In this case, previous research has shown that gender is a moderator for M1 but not for M2.
Thank you so much for this tutorial! What do I do, if the coefficients are above 1? Do I have an error in my set of data? The p value is
Coefficients are not standardized, so they can go over 1, just depends on the scale of the predictors.
Hi, how can we get standardized coefficients ?
Hi, how to check the significance of indirect effect, as there is no p, Kindly tell me.
Generally, people check the bootstrapped confidence interval. If it contains zero, you would say no mediation occurred, while if it does not contain zero, mediation has occurred.
Statistics of DOOM thnks
This video is incredibly helpful, thank you! Do you have a sample write-up of how you would report these results in APA format available? Additionally, do you have any resources on how to interpret Model 6 when one's outcome variable is categorical?
There are some examples on the OSF page, but I don't think I have one for double mediation specifically. If the DV is categorical, you should be using logistic regression (as linear regression assumes a scale DV) rather than process.
Statistics of DOOM Thanks! When I ran the analysis with process the output provided was that of a logistic regression - is that going to be an erroneous result? If so, any resources you know of for performing serial mediation with a categorical DV without using process? Thanks again!
Maybe that's been added to the newest process (I am unsure but Hayes website would say) - it wouldn't surprise me if it had been added. Otherwise, I would suggest just running the same regressions yourself one step at a time using the logistic regression options (that would also allow you to check the output you have). The problem would be the bootstrapping mediation effect - and I am unsure a good way to do that that doesn't involve using the paid bootstrapping plug in or R.
Great mediation video, I feel so relaxed that I'm going to sell all my weed, thank you, so much, let's be RUclips pals?
Hi Erin, is it necessary to have a meaningful relationship between the mediators variables ? Thank you so much
Depends on if you think you have two separate mediators (M1 does not predict M2 but X->M1->Y and X->M2->Y) or you think you have serial mediation X -> M1 -> M2 -> Y. I would think that would be based on theory, which way you would expect the relationships to pan out.
Thank you very much Dr. Erin
Hello again Dr. Erin. According to theory M1 -> M2 must be related. Already a bivariate relationship .68*. But while serial mediation M1 and M2 are not significantly related, X -> M1 -> M2 -> Y indirect relationship is significant according to confidence intervals. Can this result be interpreted as a serial mediator? The total effect and the total indirect effect are significant, while the direct effect is based on meaningless confidence intervals. Thank you very much for your support
I would think that if the indirect effect CI does not include zero, you could say that it is mediated, but that M1->M2 did not appear in this sample.
Thank you very much :)
Could you please cover M4 and M40, are so important
Thanks
I have a video for model 4 - search for single mediation with process.
Please how do you plot moderated-mediation for model 7
I don't know that you would do a plot for model 7, since the focus is the mediation.
Hi professor this was so clear to me to make my master thesis analysis, in this part of my thesis your videos was a coadviser to me, i have a basic question that ı couldnt get a clear answer i made a serial mediation (model 6) like that and their order has a theoretical order, they both significant mediators but when i writing the results i couldn't get a proper APA format (i checked the stat tools ) could you please help me
Hi, I'm not sure what you are asking? An example write up? I would find a previous publication and mimic what they did honestly.
Thank you SO much
This really very informative tutorial. i have a question, can u plz specify on which conditions we can say its a serial mediation or double mediation??
all three indirect effects (Ind1, Ind2, Ind3) should be significant?? in my model i got
Ind: X---->M1----->Y insignifcant.
Ind2: X--->M1--->M2-->Y sig
Ind:3 X-->M2-->Y sig
kindly help me to consider it a full double mediation while direct effect of X on Y is also insignifcant.
That would depend on the index of mediation provided in the output - sounds like double mediation with the sig paths, but hard to know with out the confidence interval on the indirect effects.
@@pnaracet1562 I would just describe it in your paper as you have here. Sounds like M1 mediation is happening, M2 mediation is happening but not M1 --> M2 since that's usually ind3. So, two single mediations.
Is this linear double
Not sure what you are asking - this is "double" mediation, sometimes called serial mediation with two mediators.
Great video, but flipping between screens is very distracting and disorienting
Hello. Thank you for your videos, they are brilliant!
I am using PROCESS for the first time. When I run two separate mediation analysis (X-->M1-->Y and X-->M2-->Y) both mediators are significant. However, when I run parallel or serial mediation, my mediators are not significant. Can anyone please help me understand why this may be?
Sounds like suppression to me ... M1 and M2 might be very highly correlated?
Thank you for your response!! This is certainly the case. M1 and M2 are very highly correlated. Please could you tell me what the appropriate way to deal with suppression would be?
If M1 and M2 are that highly correlated, then that implies that they are essentially the same variable (like so interrelated, measuring one of them is enough). I'd either pick one or do two separate mediations and show how they give you the same results.
Thank you so much for your help! :D
Wish to see your cat!!!
Ha! They are super interested in the computer only when recording. I'll try to catch them at some point.
Hello Ma'am
Excuse me Dr. Erin. I have sent you an email. Could you please check your inbox out and see my research model with regression report and help me understand it? Thanks in advance!
Great video! On which theoretical literature did your tests for outliers (Mahalanobis), normality, linearity, homoscedasticity base? Is there literature I can cite for my master thesis?
The Tabachnick and Fidell Multivariate statistics book!