PROCESS accepts a dichotomous Y variable. If the entered variable for Y is identified as dichotomous, PROCESS will employ logistic regression to model the Y variable. The coefficients in the output are expressed in a log-odds metric. Instead of the typical R-squared, McFadden's R-squared, Cox and Snell's R-squared, and Nagelkerke's R-squared are reported.
@@AppliedStats Thank you for your response. I conducted a mediation analysis in PROCESS with a binary DV, a binary IV, and a continuous mediator. My findings are as follows: 1) The first model (X -> M) was significant. 2) In the second model, the X (IV) was not a significant predictor of the Y (DV) but the M (mediator) was. 3) In the third model, the direct effect of X on Y was not significant. The indirect effect of X on Y was significant. What do you make of it? I think this means that the mediator completely explains the relationship between the IV and the DV. So I have a complete mediation, no?
@@AppliedStats One more thing. The same IV significantly predicts the DV if I run a binary logistic regression. Does it then make sense that the same IV does not predict the same DV in mediation analysis? In PROCESS: 1)The direct effect of X on Y is the path which doesn't include the mediator. 2) The indirect effect of X on Y is the path which includes the mediator. Do I understand it correctly?
I explain in the video where to get the direct effect and where to get the indirect effect. the third regression in the output does not give you the direct effect (c'), it gives you the total effect (c).
Thank you for the video! I just confused with my results and I hope that you could help me.. According to my results of Mediation Analysis (Model 4) there is significant mediation and indirect effect. However, sign of coefficient is a bit confusing. Here there are: a=.70, b=.39, c'= .51. Since my outcome variable is binary (categorical) I had really difficulties to explain and interpret the result accordingly? Could you please help me with this? (Predictor: Agreeableness, Outcome: Sustainable Investment Intention, Mediator : Green Consumption Attitude)
Hello, thanks for your video! It is super helpful! I want to ask one thing, in your report, [the bootstrap resamples were above entirre 0 (from 0.049 to 0.197)]. I cannot find 0.197 from your spss output window, how did you calculate this? Thank you in advance :)
Good question. Bootstrapping is a resampling technique that involves drawing random samples from the original sample. Each time we bootstrap, we may end up with slightly different confidence intervals due to the random nature of the resampling process. It appears that my write-up in the Word file was based on a previous run of the analysis. In the output, shown in the video the interval is .049-.201, which is very close to the 0.049 to 0.197 range you can see in the Word file.
@@AppliedStats Thank you so much! I have another question, from my analysis my c' path p value is 0.913. (path a and b's p value are .019 and < .001 respectively). In this case, can I interpret my model is significant? (indirect effect is significant though)
If you have a negative indirect effect bootllci and a positive bootulci does that still mean that your mediator is significant, what does a negative confidence interval represent? Does this represent a negative relation?
Hello, thank you for a very helpful video. I had a question I was hoping you could help with. I am doing a process macro analysis (model 6) where Y: Burnout, X: Family dynamic, with psychological capital and social support as mediators. I wanted to add ethnicity as a possible mediator as well, since the data shows that those of a different ethnicity than white have better family dynamics and so experience less burnout. However, I received an error saying that I cannot have a dichotomous variable as a mediator. Is there another way for me to include ethnicity somehow? Would appreciate any assistance you can provide.
PROCESS does not allow dichotomous mediators. I am not aware of a workaround. But I also do not think that ethnicity can be used as a mediator anyways.
If I understand your question correctly, your statistical hypothesis may be: No mediation: The mediator has no effect, the predictor only directly affects the outcome. Mediation: The mediator explains part of the effect of the predictor on the outcome.
This saved my thesis. Thank you so much
Would the interpretation be the same when the DV is binary? None of the available videos actually interpret such mediations.
PROCESS accepts a dichotomous Y variable. If the entered variable for Y is identified as dichotomous, PROCESS will employ logistic regression to model the Y variable. The coefficients in the output are expressed in a log-odds metric. Instead of the typical R-squared, McFadden's R-squared, Cox and Snell's R-squared, and Nagelkerke's R-squared are reported.
@@AppliedStats Thank you for your response. I conducted a mediation analysis in PROCESS with a binary DV, a binary IV, and a continuous mediator. My findings are as follows: 1) The first model (X -> M) was significant. 2) In the second model, the X (IV) was not a significant predictor of the Y (DV) but the M (mediator) was. 3) In the third model, the direct effect of X on Y was not significant. The indirect effect of X on Y was significant. What do you make of it? I think this means that the mediator completely explains the relationship between the IV and the DV. So I have a complete mediation, no?
@@AppliedStats One more thing. The same IV significantly predicts the DV if I run a binary logistic regression. Does it then make sense that the same IV does not predict the same DV in mediation analysis?
In PROCESS:
1)The direct effect of X on Y is the path which doesn't include the mediator.
2) The indirect effect of X on Y is the path which includes the mediator.
Do I understand it correctly?
If the direct effect is not significant and the indirect effect is significant, you have full or complete mediation.
I explain in the video where to get the direct effect and where to get the indirect effect. the third regression in the output does not give you the direct effect (c'), it gives you the total effect (c).
Thank you for the video! I just confused with my results and I hope that you could help me.. According to my results of Mediation Analysis (Model 4) there is significant mediation and indirect effect. However, sign of coefficient is a bit confusing. Here there are: a=.70, b=.39, c'= .51. Since my outcome variable is binary (categorical) I had really difficulties to explain and interpret the result accordingly? Could you please help me with this? (Predictor: Agreeableness, Outcome: Sustainable Investment Intention, Mediator : Green Consumption Attitude)
Thank u sir .
Hello, thanks for your video! It is super helpful! I want to ask one thing, in your report, [the bootstrap resamples were above entirre 0 (from 0.049 to 0.197)]. I cannot find 0.197 from your spss output window, how did you calculate this? Thank you in advance :)
Good question. Bootstrapping is a resampling technique that involves drawing random samples from the original sample. Each time we bootstrap, we may end up with slightly different confidence intervals due to the random nature of the resampling process. It appears that my write-up in the Word file was based on a previous run of the analysis. In the output, shown in the video the interval is .049-.201, which is very close to the 0.049 to 0.197 range you can see in the Word file.
@@AppliedStats Thank you so much! I have another question, from my analysis my c' path p value is 0.913. (path a and b's p value are .019 and < .001 respectively). In this case, can I interpret my model is significant? (indirect effect is significant though)
if the ci for indirect effect does not include zero, you have a significant mediation
If you have a negative indirect effect bootllci and a positive bootulci does that still mean that your mediator is significant, what does a negative confidence interval represent? Does this represent a negative relation?
this means your indirect effect is not significant
@@AppliedStats Yes, thanks.
Hello, thank you for a very helpful video. I had a question I was hoping you could help with. I am doing a process macro analysis (model 6) where Y: Burnout, X: Family dynamic, with psychological capital and social support as mediators. I wanted to add ethnicity as a possible mediator as well, since the data shows that those of a different ethnicity than white have better family dynamics and so experience less burnout. However, I received an error saying that I cannot have a dichotomous variable as a mediator. Is there another way for me to include ethnicity somehow? Would appreciate any assistance you can provide.
PROCESS does not allow dichotomous mediators. I am not aware of a workaround. But I also do not think that ethnicity can be used as a mediator anyways.
@@AppliedStats okay, thank you for the reply!
Awesome thank you!
How do you formulate the statistic hypothesis for the mediation?
If I understand your question correctly, your statistical hypothesis may be:
No mediation: The mediator has no effect, the predictor only directly affects the outcome.
Mediation: The mediator explains part of the effect of the predictor on the outcome.
@@AppliedStats Thank you but I mean like with regression:
H0: β = 0
H1: β < 0
-> how to do this with a mediation?
H0: ab = 0
H1: ab ≠ 0
@@AppliedStats
Thank you very much!
But what does "ab" stand for?
From the video, ab = indirect effect which explains whether there is a mediation or not. I hope I’m correct?
Could you offer a word download of the table you use?
is total effect of no use in mediation analysis?
Knowing the total effect can be useful. it is the sum of direct and indirect effects. it would not hurt to report it.
How to analyze with numerous factors ?
you can send more than 1 mediator to the mediator box