Hi! I have a question... do we need to to a linear regression first (or any other analysis first) before we do the medmod moderation analysis or we can do the moderation analysis on its own?
Awesome! Thanks for the Video! Did i get this right that you use the medmod package (which is like a mini-sem) as some kind of post-hoc for the significant moderation in the second regression model?
Hey Rene, thank you for the tutorial. It helped me a lot. However, I have one issue with the estimates of the simple slope analysis: when the interaction did not have significant p-values in the moderation model, how come at the "average" level of the moderator jamovi computes a significant value for the interaction? Shouldn't that value also be not significant at this level (average)? I am having trouble to understand how I can report the output correctly. Again - thank you for you great tutorials!
When the interaction is not significat, reporting the simple slopes may not be necessary. A non-significant interaction means that the realtionship between the IV and DV (regardless if it is significant or not significant) will be the same across all levels of the moderator. Any variarion in the significance across levels of the moderator might be acticfactual.
Hey, thanks for the video. I a have a question : why the bullying effect is not significative when you use the regression module, but significative when you use the moderation effect modulle ? I have the same problem with my variables.
As I have explained in the video, the table generated in the moderation module of Jamovi is basically two tables combined. The main effects of bullying and peer support is from a separate calculation with only those two variables enter as predictors. On the other hand, the output from the interaction term is from a different calculation with bullying, support, and interaction entered as predictor variables. In the multiple regression model with bullying, support, and interaction entered as predictor variables, bullying is no longer a significant variable because the coefficients displayed in now under the assumption the support and the interaction is the same or constant for everyone. However, you should not be interpreting this. the only relevant result in the particular model in the interaction effect. The main effects to be interpreted is the one in the model that does not include the interaction. I hope I made some sense. thanks.
@@renenob Thanks ! So, just to be sure, we can say that bullying has a significative effect on anger, and that the relation between those variables is significantly moderated by Peer Support ?
Rene thanks for your videos, very good. You mention that for a categorical moderation (Gender in my case) on a continuous DV you would use a different way of calculating moderation and refer to a previous video. Could you please link me to the video? I searched on your channel and there is too much for me to know which one it is. Also, might a MEDIATION analysis be ok in the case of a categorical variable like Gender? Thank you.
Sorry for the extreme late reply. Gender is unlike to be a mediator. A moderator is essentially (also) a DV of the original IV. Gender (sex) is unlike to be influenced by anything other the specific combination of sex chromosomes. As for categorical moderators. I recommend that you look for tutorials on Hayes Process Macro. it can handle dichotomous moderators.
I encountered this one, thank you Sir Nob!
welcome!
Hi! I have a question... do we need to to a linear regression first (or any other analysis first) before we do the medmod moderation analysis or we can do the moderation analysis on its own?
Thank you very much for the tutorial.
What if my moderation variable is not linear?
what do you mean? Linear usually refers to the quality of the relationship between 2 variables, not just a variable.
Hey, thanks for the video! I also do have a question: If using medmod, can the predictor be dichotom?
I think so
Awesome! Thanks for the Video! Did i get this right that you use the medmod package (which is like a mini-sem) as some kind of post-hoc for the significant moderation in the second regression model?
Hey Rene, thank you for the tutorial. It helped me a lot.
However, I have one issue with the estimates of the simple slope analysis: when the interaction did not have significant p-values in the moderation model, how come at the "average" level of the moderator jamovi computes a significant value for the interaction? Shouldn't that value also be not significant at this level (average)?
I am having trouble to understand how I can report the output correctly.
Again - thank you for you great tutorials!
When the interaction is not significat, reporting the simple slopes may not be necessary. A non-significant interaction means that the realtionship between the IV and DV (regardless if it is significant or not significant) will be the same across all levels of the moderator. Any variarion in the significance across levels of the moderator might be acticfactual.
Hey, thanks for the video. I a have a question : why the bullying effect is not significative when you use the regression module, but significative when you use the moderation effect modulle ? I have the same problem with my variables.
As I have explained in the video, the table generated in the moderation module of Jamovi is basically two tables combined. The main effects of bullying and peer support is from a separate calculation with only those two variables enter as predictors. On the other hand, the output from the interaction term is from a different calculation with bullying, support, and interaction entered as predictor variables. In the multiple regression model with bullying, support, and interaction entered as predictor variables, bullying is no longer a significant variable because the coefficients displayed in now under the assumption the support and the interaction is the same or constant for everyone. However, you should not be interpreting this. the only relevant result in the particular model in the interaction effect. The main effects to be interpreted is the one in the model that does not include the interaction. I hope I made some sense. thanks.
@@renenob Thanks ! So, just to be sure, we can say that bullying has a significative effect on anger, and that the relation between those variables is significantly moderated by Peer Support ?
Yes
Rene thanks for your videos, very good. You mention that for a categorical moderation (Gender in my case) on a continuous DV you would use a different way of calculating moderation and refer to a previous video. Could you please link me to the video? I searched on your channel and there is too much for me to know which one it is. Also, might a MEDIATION analysis be ok in the case of a categorical variable like Gender? Thank you.
Sorry for the extreme late reply. Gender is unlike to be a mediator. A moderator is essentially (also) a DV of the original IV. Gender (sex) is unlike to be influenced by anything other the specific combination of sex chromosomes. As for categorical moderators. I recommend that you look for tutorials on Hayes Process Macro. it can handle dichotomous moderators.
Gracias