@@researchwithfawad Am deeply appreciative for your explanation. I went over the video and output many times and could not come up with where the t-values were coming from. These videos are so clear and helpful.
I mostly was able to recreate the table and narrative of the Mediation Analysis. However, I was quite at a loss since I don't know where to find the t-statistic equivalent. May I know where I should or how I should look for it?
@@researchwithfawad hello, sorry but I don't understand for the calculation. Can you explain me with the numbers ? Thanks a lot for your video very useful !
Sorry sir, I've got one more question. Why the results of BootLLCI and BootULCI are different every time when I run the program even all command button remains the same? Thank you!
Thanks for your query. This is because you are using Hayes Process Macro that is estimating results based on the bootstrapping technique that uses random number generation.
@ResearchWithFawad, thanks for the video, excellent! I have a question about the interpretation of the results: what do you mean by there being no zero between the lower and upper levels. could you please elaborate on that? where would I see a zero if the effects are not significant?
Thank you for the video - very informative! I have a question: How do I calculate the effect size (preferably R2) for indirect effects? PROCESS output shows effect size (R2) for all effects except indirect. Any help would be appreciated!
Pleasure. The concept of R sq is to explain impact of exogenous variable(s) on the dependent variable. For indirect effects focus on beta and type of Mediation
Hi, how if one mediating variable (let's say M1) is not significant? So the conclusion would be no mediation through M1? And partial mediation through M2?
thank you for your very informative video. I have a question: What is the difference between Model 4 with multiple medaitors and Multiple mediation using bootstrapping in SPSS?
A.o.a sir sorry for bothering u again I have one more question is there a possibility that direct and indirect effects are significant but the total effects aren't significant? If yes, then does it mean that there was no mediation even though indirect effects were significant
It is unlikely to happen. As the path making up the total effects are significant so how can the sum of it be insignificant of individual paths are significant.
I have run parallel mediation analysis total effects were non significant whereas direct and indirect effects through 2 mediators out of the three mediators were significant
Thank you for your video, it is really helpful. I would like to ask one question. In my model, I have 3 mediators which are related concepts with each other. So, should I conduct mediation Process Macro 4 for each mediator separately or together? When I put them separately, I found significant mediation for each mediator, but together 1 is not. Any help would be appreciated!
When I run my mediation analysis with 4 mediators together, when 3 mediation results were not significant. But separately, I have significant results. My professor insists on parallel mediation analysis, because she thinks if there is a relationship between IV and DV, we can't say that: it is just because of the simple mediator. Se told that If I thought all of them can effect the relationship between IV and DV, I should keep them together. She recommended me to apply VIF analysis but my score in there again minus 2, I think this problem is not about VIF either. If you have possible solution for solving this problem, could you please help me?
Thank you for the video! It was really informative and helped me a lot. I have 2 questions though: 1) If I want to compare which path (a or b) has a stronger impact on DV, then I just have to test if the indirect effects are stat different and compare them, right? 2) In my case I have two IVs. Can I use Hayes Process for that? Thanks a lot!
Much appreciated. But what about the whole indirect effect? For example the IV->M1->DV and IV->M2->DV. Can they be compared? I want to examine whether a1*b1 (Indirect effect 1) or a2*b2 (Indirect effect 2) is more efficient in increasing the DV. @@researchwithfawad
If the LL and UL are in opposite directions, that is one is positive and other is negative this means there is a 0 between the two. Meaning, the hypothesis is not supported.
thank you so much - this is so informative I do have a question about power calculations - so how can we conduct power analysis for multiple mediators? Thank you!
Thanks for your appreciation. You can refer to this paper Jak, S., Jorgensen, T. D., Verdam, M. G. E., Oort, F. J., & Elffers, L. (2020). Analytical power calculations for structural equation modeling: A tutorial and Shiny app. Behavior Research Methods. doi.org/10.3758/s13428-020-01479-0
Thanks a lot for such an amazing video. I have a question sir. a variables with subscales (e.g.,3 subscales) and fow which separate scores of subscales have been calculated no cumulative score has been calculated can be used as a mediator in parallel mediation?If yes ,do we have to add another variable ?
@@researchwithfawad Sir I have a question... in my analysis I have used two mediators and there is one IV and DV. I am getting significant a1,a2,b1,b2. my indirect effect and total effect is also significant. but direct effect between IV and DV in not significant. Is this mediation?
Also, I am not sure what is the meaning of coefficient and standardized coefficients. I want to know is positive or negative impact between two values.
Thanks for your illustration sir, it is very helpful. However, if my BootLLCI is -0.0002 and Boot .1062 in the indirect effect of X on Y, is this means the mediation is not significant? Thanks!
Thanks. I am glad you liked it. Well, have you tried to also calculate t-statistics, Divide Estimate by BootSE. Just to add more support to your mediation results. see if the value is > 1.96.
If the direct effect is insignificant and the indirect effect is significant then it is full mediation, if both effects are significant, it is partial, if only direct effect is significant, it is no mediation.
Sir, I would like to express my gratitude for you because this video and, of course, your explanation really help me to understand the analysis. I have some questions regarding my results. I found that my A1, B1, and C' are not significant, while A2 and B2 are. The total effect is significant tho. However, there is no mediation found because the indirect effects are not significant as well. How do I interpret this one? I mean, how do I explain that the effects of X on M2 is significant, M2 on Y is significant as well, but there's no mediation? I really have no idea about this. I apologize if I asked too much, but thank you for this opportunity. 🙏
Meanwhile, my other model found another result which I could understand a little bit better. The A1, A2, and B2 are significant, while C' and B1 are not, but it seems like there's a full mediation for the path 2 so I would assume that the effects of X on Y is fully mediated by M2, and M1 is not a mediator. Is that correct, Sir? I appreciate any feedback. Thank you.
Pleasure. You will need to find reasons as to what the relationships could be insignificant. What are possible reason for contradictions. This video may also help ruclips.net/video/xPZymvNG-gk/видео.html
Hello Sir. I am really sorry that this question is not specific to this video. Can you please mention the values other than Cronbach alpha values which must be shown to support a pilot study? Also, what is the best way to estimate the sample for a pilot study? ( I plan to work with around 370 respondents, so can I conduct a pilot study with 50 respondents?) The kind of confidence your videos give us, really feel like to get every aspect of research known from you.Thanks a lot Sir.
Thanks for your comment. I am happy that the video are helpful. As for sample size, normally 50 is enough. Other than Cronbach Alpha, report if there were any changes made to the wording of the items based on the input of the respondents in the pilot study.
In addition to my earlier question, I would like to mention that the cronbach alpha values came out perfect. So can I consider those responses as a part of my total responses? Or do i have to exclude these responses (responses from pilot data) from my final survey data?
If adding two mediators simultaneously (parallel mediation) makes the indirect effect non significant (when it was significant when simple mediation was used) how can this be interpreted? The sign for the indirect effect is the same for both mediators.
@@researchwithfawad interesting. Even if they have the same sign? 2) but our initial single mediator model still holds true right? From parallel mediation, we just know that in the presence of the other one, it doesnt matter.
The interest is in significance, not the sign at the moment. Yes, if you are testing one single mediator, the relationship is significant, see if the direct effect also changes with inclusion or exclusion of mediators.
@@researchwithfawad For example, de p value in "Direct effect of X on Y" is .750 so is not significant. But the BootLLCI is -2.6148 and BootULCI is -.2033. I have a mediation in this case?
AOA. If there are two mediators (but from the same scale i.e two "conditions" or aspects of the same one variable) would we do mediation analysis like you showed in this video OR would we carry out simple mediation analysis putting both of the conditions one by one etc. P.S the two conditions of the mediating variable are also correlated with each other.
WAS. It depends, if you are assessing each dimension aas an individual intervening factor, then you will need to have separate mediating hypothesis for each sub dimension. Otherwise you can have it as composite
@@researchwithfawad OK, so I am assessing each dimension as in individual intervening factor. If I carry out the analysis as you did in this video, will the results be comparable to how it would be if say, I checked mediation by putting in one mediating variable AT A TIME (as opposed to putting in the two simultaneously as showed in this video eg IV, M1,M2, DV in one analysis), The former would mean I would need to carry out two separate mediation analysis (IV, M1, DV and IV, M2, DV) but the results would be the same right?
No, do not do separate mediation analysis for each factor, put each intervening variable separately in a single analysis, this is much more meaningful and closer to achieving parsimony.
@@researchwithfawad Another question in response to your comment. So, in Hayes book on mediation and moderation, the authors say "in a parallel mediator model, antecedent variable X is modeled as influencing consequent Y directly and indirectly through two or more mediators with the condition that no mediator causally influences another”. Does that mean this condition NECESSARY IF we want to carry out a parallel mediation model or can we run It without it as well? Because like you said this analysis is more parsimonious, so can I still run it if my two mediators are connected to one another (because they are as they are from the same theory and broader construct so to say).
What If we had negative total effect and indirect effects had negative results too , direct effect would be positive after that, How it can be interpret , because in such a situation , there would be such as total effect,. . -10 and direct effect .13 .. I wonder how it can be interpret in such a complex situation
Results can be positive or negative. If they are in line with the proposed hypotheses, there shouldn't be any issue. For example Stress as IV can negatively affect the mediator (Commitment) and that can positively affect Job performance. However, the IV can negatively affect Performance (DV)
AOA. how do we interpret results if indirect effect is significant for M1, but non significant for M2 AND direct effect is significant? is there still mediation? Will it be for M1? 2) what if if indirect effect is significant for M1, but non significant for M2 AND direct effect is nonsignificant. What is it non significant for?
Yes, Mediation for M1 but not M2. Partial mediation, if direct effect is significant. But when Direct effect is insignificant, it means full mediation, no matter which mediation path one is significant or insignificant.
Also if the direct effect is non significant and so is the indirect effect, how do we report it in text i.e what could our statement look like? "MI and M2 didn't mediate the relationship b/w X and y"?
If the direct effect is insignificant and the indirect effects are insignificant, this mean no mediation. The M1 and M2 does not have any mediating role in explaining the impact of IV on DV.
Beautifully explained the entire matter. Big Thank you Mr. Fawad....🙏
Thanks for watching.
It was one of the very easy lectures I found on RUclips. Well done👍
Thanks for watching.
you're my hero for my dissertation mannn!!
Thanks for watching
Thank you for your video. It’s helpful 😊
Thanks for watching
Very lucid and clear explanation. kindly tell me from where you found the t-statistic value of 4.068 and 4.362 in Table 1.
Thanks. I am glad you liked it.
Kindly divide the Estimate by Standard Error to get t value.
@@researchwithfawad Am deeply appreciative for your explanation. I went over the video and output many times and could not come up with where the t-values were coming from. These videos are so clear and helpful.
I mostly was able to recreate the table and narrative of the Mediation Analysis. However, I was quite at a loss since I don't know where to find the t-statistic equivalent. May I know where I should or how I should look for it?
Thanks for your comment. To get t statistics divide the estimate by standard error
@@researchwithfawad May I know where the SD is found in 9:30? 🤔
You will find it in the indirect effect. BootSE
thank u for this video, although i was confused about the t test value, how u hot it as its nor appeared in the findings
Pleasure. I am glad you liked it. To get t statistics. Divide the bootstrap estimate by Boot Standard Error (SE)
Thank you, prof. I have 2 questions:
1- why you mentioned p
Pleasure.
When p value is 0.000, we can report it like it.
You can report other statistics as well.
Great video thanks. Where did you get the indirect effects t values from? I cannot see them in your output.
Thanks. I am glad you liked it. You can divide the Estimate (IndirectEffect) by Standard Error (SE).
selamunaleykum, how do you calculate t values (as 4.068, 4.362) on table end of video?
WAS. Please divide Path Coefficient by SE
@@researchwithfawad hello, sorry but I don't understand for the calculation. Can you explain me with the numbers ? Thanks a lot for your video very useful !
Sorry sir, I've got one more question. Why the results of BootLLCI and BootULCI are different every time when I run the program even all command button remains the same? Thank you!
Thanks for your query. This is because you are using Hayes Process Macro that is estimating results based on the bootstrapping technique that uses random number generation.
A.o.a Sir, for parallel mediation do we have to add mediators separately or together in the mediators column and would the Model 4 be used
Please watch the video.
@ResearchWithFawad, thanks for the video, excellent! I have a question about the interpretation of the results: what do you mean by there being no zero between the lower and upper levels. could you please elaborate on that? where would I see a zero if the effects are not significant?
Thanks for watching. Check if there is a0 between the LL and UL CI, if not then the proposed relationship is significant.
Thank you for the video - very informative! I have a question: How do I calculate the effect size (preferably R2) for indirect effects? PROCESS output shows effect size (R2) for all effects except indirect. Any help would be appreciated!
Pleasure. The concept of R sq is to explain impact of exogenous variable(s) on the dependent variable. For indirect effects focus on beta and type of Mediation
thank you this was very helpful :) i have a question. how did you create the graphic model for the mediation analysis?
Thanks for your interest. I am glad you liked it. You can do it in Power Point or Ms Word.
Hi, how if one mediating variable (let's say M1) is not significant?
So the conclusion would be no mediation through M1? And partial mediation through M2?
Each mediator is assessed separately. Other may be significant.
thank you for your very informative video. I have a question: What is the difference between Model 4 with multiple medaitors and Multiple mediation using bootstrapping in SPSS?
Multiple mediation in SPSS? Explain
A.o.a sir sorry for bothering u again I have one more question is there a possibility that direct and indirect effects are significant but the total effects aren't significant? If yes, then does it mean that there was no mediation even though indirect effects were significant
It is unlikely to happen. As the path making up the total effects are significant so how can the sum of it be insignificant of individual paths are significant.
I have run parallel mediation analysis total effects were non significant whereas direct and indirect effects through 2 mediators out of the three mediators were significant
Hey. How did you get t stat value in the table?
Thanks for watching. You can divide the estimate by Standard Error to get t statistics.
Thank you for your video, it is really helpful. I would like to ask one question. In my model, I have 3 mediators which are related concepts with each other. So, should I conduct mediation Process Macro 4 for each mediator separately or together? When I put them separately, I found significant mediation for each mediator, but together 1 is not. Any help would be appreciated!
Pleasure. I am glad you liked it. Model 4 with separate mediators is appropriate.
@@researchwithfawad thank you for your kindly support.
When I run my mediation analysis with 4 mediators together, when 3 mediation results were not significant. But separately, I have significant results. My professor insists on parallel mediation analysis, because she thinks if there is a relationship between IV and DV, we can't say that: it is just because of the simple mediator. Se told that If I thought all of them can effect the relationship between IV and DV, I should keep them together. She recommended me to apply VIF analysis but my score in there again minus 2, I think this problem is not about VIF either. If you have possible solution for solving this problem, could you please help me?
This happens when there are multiple relationships assessed simultaneously. You can increase your sample size and see if there is a change in results.
Thank you for the video! It was really informative and helped me a lot. I have 2 questions though:
1) If I want to compare which path (a or b) has a stronger impact on DV, then I just have to test if the indirect effects are stat different and compare them, right?
2) In my case I have two IVs. Can I use Hayes Process for that? Thanks a lot!
Thanks for watching
Path a is from IV to MV notnDV so there cannot be a comparison.
You cannot use Hayes with two IV. For this use SmartPLS or AMOS.
Much appreciated. But what about the whole indirect effect? For example the IV->M1->DV and IV->M2->DV. Can they be compared? I want to examine whether a1*b1 (Indirect effect 1) or a2*b2 (Indirect effect 2) is more efficient in increasing the DV. @@researchwithfawad
what does there is no zero in between class intervals mean? Is it about 0 before or after decimals? thank you
If the LL and UL are in opposite directions, that is one is positive and other is negative this means there is a 0 between the two. Meaning, the hypothesis is not supported.
@@researchwithfawad thankyou so much. You really saved my dissertation✨
thank you so much - this is so informative
I do have a question about power calculations - so how can we conduct power analysis for multiple mediators?
Thank you!
Thanks for your appreciation.
You can refer to this paper
Jak, S., Jorgensen, T. D., Verdam, M. G. E., Oort, F. J., & Elffers, L. (2020). Analytical power calculations for structural equation modeling: A tutorial and Shiny app. Behavior Research Methods. doi.org/10.3758/s13428-020-01479-0
Thanks a lot for such an amazing video. I have a question sir. a variables with subscales (e.g.,3 subscales) and fow which separate scores of subscales have been calculated no cumulative score has been calculated can be used as a mediator in parallel mediation?If yes ,do we have to add another variable ?
Thanks for watching. Yes you can use each sub-dimension as a separate construct in a mediation model.
Thank u so much for answering my query.
is it possible to have 2 output variables, 2 mediators and having one Independent variable. how can it be done
Not in Process Macro. You will need tonuse SEM tools like SmartPLS4
@@researchwithfawad Sir I have a question... in my analysis I have used two mediators and there is one IV and DV. I am getting significant a1,a2,b1,b2. my indirect effect and total effect is also significant. but direct effect between IV and DV in not significant. Is this mediation?
is there an explanation for how to conduct a mediational analysis in an experimental method with one mediating variable with two levels ?
If there are levels, use Mplus
Is the mediation analysis follow APA format?
Please refer to this video as well.
ruclips.net/video/mLKVPXwqT9o/видео.html
Also, I am not sure what is the meaning of coefficient and standardized coefficients. I want to know is positive or negative impact between two values.
variables
Thanks for your illustration sir, it is very helpful. However, if my BootLLCI is -0.0002 and Boot .1062 in the indirect effect of X on Y, is this means the mediation is not significant? Thanks!
Thanks. I am glad you liked it. Well, have you tried to also calculate t-statistics, Divide Estimate by BootSE. Just to add more support to your mediation results. see if the value is > 1.96.
@@researchwithfawad Thanks for your advice sir. When I divide effect by BootSE, the t value is 1.88. May I know what does it means? Thank you!
@@researchwithfawad Also, if the t-value is >1.96, what does it tells as well? Thank you very much!!
If you take it as 1 tailed, you can say the relationship is significant. for 1 tailed the criteria of t statistics is 1.645
@@researchwithfawad Thank you so much! But how should I interpret it? Do I have to mention I'm taking it as 1 tailed?
how to get the t statistics in summary analysis
Divide Estimate by BootSE (Standard Error).
If direct effect is insignificant so there will be full mediation? For both m1 and m2
If the direct effect is insignificant and the indirect effect is significant then it is full mediation, if both effects are significant, it is partial, if only direct effect is significant, it is no mediation.
@@researchwithfawad thank you so much sir ..s
Sir, I would like to express my gratitude for you because this video and, of course, your explanation really help me to understand the analysis.
I have some questions regarding my results. I found that my A1, B1, and C' are not significant, while A2 and B2 are. The total effect is significant tho. However, there is no mediation found because the indirect effects are not significant as well. How do I interpret this one? I mean, how do I explain that the effects of X on M2 is significant, M2 on Y is significant as well, but there's no mediation?
I really have no idea about this. I apologize if I asked too much, but thank you for this opportunity. 🙏
Meanwhile, my other model found another result which I could understand a little bit better. The A1, A2, and B2 are significant, while C' and B1 are not, but it seems like there's a full mediation for the path 2 so I would assume that the effects of X on Y is fully mediated by M2, and M1 is not a mediator. Is that correct, Sir? I appreciate any feedback. Thank you.
Pleasure. You will need to find reasons as to what the relationships could be insignificant. What are possible reason for contradictions.
This video may also help
ruclips.net/video/xPZymvNG-gk/видео.html
@@researchwithfawad thank you very much, I will check the video 😉
Hello Sir. I am really sorry that this question is not specific to this video. Can you please mention the values other than Cronbach alpha values which must be shown to support a pilot study? Also, what is the best way to estimate the sample for a pilot study? ( I plan to work with around 370 respondents, so can I conduct a pilot study with 50 respondents?)
The kind of confidence your videos give us, really feel like to get every aspect of research known from you.Thanks a lot Sir.
Thanks for your comment. I am happy that the video are helpful.
As for sample size, normally 50 is enough.
Other than Cronbach Alpha, report if there were any changes made to the wording of the items based on the input of the respondents in the pilot study.
Noted Sir. Thank You so much Sir for your kind reply.
In addition to my earlier question, I would like to mention that the cronbach alpha values came out perfect. So can I consider those responses as a part of my total responses? Or do i have to exclude these responses (responses from pilot data) from my final survey data?
Better to exclude them.
Thank You Sir
If adding two mediators simultaneously (parallel mediation) makes the indirect effect non significant (when it was significant when simple mediation was used) how can this be interpreted? The sign for the indirect effect is the same for both mediators.
Adding additional mediator is cancelling the effect of the other Mediator in explaining the intervening role of the first mediator
@@researchwithfawad interesting. Even if they have the same sign? 2) but our initial single mediator model still holds true right? From parallel mediation, we just know that in the presence of the other one, it doesnt matter.
The interest is in significance, not the sign at the moment. Yes, if you are testing one single mediator, the relationship is significant, see if the direct effect also changes with inclusion or exclusion of mediators.
Briefly, what values must we pay attention to in order for mediation to be present?
Please check for Specific Indirect effects.
@@researchwithfawad For example, de p value in "Direct effect of X on Y" is .750 so is not significant. But the BootLLCI is -2.6148 and BootULCI is -.2033. I have a mediation in this case?
If t statistics values are less than 1.9 so how can interpretate?
This means that the relationship is insignificant. You will have to explain the reasons in discussion section.
Can we send our interpretation file to you to check plz
Yes, if it is just a small read. Email me on
kh.fawad83@gmail.com
If there is zero in between so indirect effect will b insignificant? And if yes then mediation will exist?
AOA. If there are two mediators (but from the same scale i.e two "conditions" or aspects of the same one variable) would we do mediation analysis like you showed in this video OR would we carry out simple mediation analysis putting both of the conditions one by one etc. P.S the two conditions of the mediating variable are also correlated with each other.
WAS. It depends, if you are assessing each dimension aas an individual intervening factor, then you will need to have separate mediating hypothesis for each sub dimension. Otherwise you can have it as composite
@@researchwithfawad OK, so I am assessing each dimension as in individual intervening factor. If I carry out the analysis as you did in this video, will the results be comparable to how it would be if say, I checked mediation by putting in one mediating variable AT A TIME (as opposed to putting in the two simultaneously as showed in this video eg IV, M1,M2, DV in one analysis), The former would mean I would need to carry out two separate mediation analysis (IV, M1, DV and IV, M2, DV) but the results would be the same right?
No, do not do separate mediation analysis for each factor, put each intervening variable separately in a single analysis, this is much more meaningful and closer to achieving parsimony.
@@researchwithfawad Another question in response to your comment. So, in Hayes book on mediation and moderation, the authors say "in a parallel mediator model, antecedent variable X is modeled as influencing consequent Y directly and indirectly through two or more mediators with the condition that no mediator causally influences another”. Does that mean this condition NECESSARY IF we want to carry out a parallel mediation model or can we run It without it as well? Because like you said this analysis is more parsimonious, so can I still run it if my two mediators are connected to one another (because they are as they are from the same theory and broader construct so to say).
Could you do the same analysis for 4 mediators?
Yes, you can. You can have up to 10 mediators.
@@researchwithfawad Thank you :)
What If we had negative total effect and indirect effects had negative results too , direct effect would be positive after that, How it can be interpret , because in such a situation , there would be such as total effect,. . -10 and direct effect .13 .. I wonder how it can be interpret in such a complex situation
Results can be positive or negative. If they are in line with the proposed hypotheses, there shouldn't be any issue.
For example
Stress as IV can negatively affect the mediator (Commitment) and that can positively affect Job performance. However, the IV can negatively affect Performance (DV)
AOA. how do we interpret results if indirect effect is significant for M1, but non significant for M2 AND direct effect is significant? is there still mediation? Will it be for M1?
2) what if if indirect effect is significant for M1, but non significant for M2 AND direct effect is nonsignificant. What is it non significant for?
Yes, Mediation for M1 but not M2. Partial mediation, if direct effect is significant.
But when Direct effect is insignificant, it means full mediation, no matter which mediation path one is significant or insignificant.
@@researchwithfawad thank you for clarifying!
Sir please bta den APA 7 k according kese result ka table bnain gy thesis k liye?
Please watch this video
ruclips.net/video/mLKVPXwqT9o/видео.html
Also if the direct effect is non significant and so is the indirect effect, how do we report it in text i.e what could our statement look like? "MI and M2 didn't mediate the relationship b/w X and y"?
I mean, is that conceptually true?
If the direct effect is insignificant and the indirect effects are insignificant, this mean no mediation. The M1 and M2 does not have any mediating role in explaining the impact of IV on DV.
@@researchwithfawad thank you very much!
How to get the value of T-statistics?🤣
Divide the Effect Size (Path Coefficient) by Standard Error.