How do I... DO A MEDIATION REGRESSION in Jamovi? (2022)

[@m.s.silver3410]

Hi, thank you for this quick introduction to medmod. Is there a way to test for assumptions for mediation analysis (heteroscedasticity or linear relation) in jamovi? Thanks!

[@AlexanderSwan]

You would need to use the linear regression module for those stats, I think.

[@m.s.silver3410]

@@AlexanderSwan Thanks!

[@aleksandrakwasnicka8765]

Dan is Danielle J. Navarro (one of the authors of LSJ) so mum, not dad.

[@AlexanderSwan]

You’re absolutely right! I recorded this and had been using the data before I realized that.

[@joshuapickup2372]

Have you got any suggestion for how to report an insignificant mediation result?

[@AlexanderSwan]

I’d say report that whatever paths are nonsig and give the p-value. It really depends on how important the mediation was to the overall narrative.

[@gelatalavera98]

Can I use nominal variable with dummy code as independent variable?

[@AlexanderSwan]

Yep!

[@elsaboulard744]

Thank youuu it was clear ! If I want to draw the model, I need to put the stand estimate to the a, b and c paths ? (sorry for my english, I'm french)

[@AlexanderSwan]

Yes correct

[@davul01]

I would like to ask you about what test to use to test the following: I have empathy pathology (scale - 0 - 4/ lowest - highest severity) measured before intervention (group therapy) and after it via a clinical interview. What test do I use to know whether the initial level of anxiety, depression and quality of life (all continuous variables 0 - 63 points) had an effect on the change in empathy pathology (did those with higher level of X change more?); more precisely it's lowering (which I have already observed using paired-sample t-test).

[@AlexanderSwan]

Sounds like you’ll wan to do a multiple regression analysis of these continuous variables to see if the anxiety, depression, and quality of life changed or predicts their empathy pathology. That’s what I understand from your comment.

[@davul01]

@@AlexanderSwan Thank you for your answer! Yea that is one way. Was just wondering if it is possible to somehow test effect of those 3 variables on the change between those two measurements (at different times before and after intervention) which were already compared using paired sample t test. I would like to know if indicial level of either of these variables was behind the change in some subject (if high level of anxiety before the therapeutic stay predicted the change in empathy pathology? Would a mediation analysis be appropriate?

[@AlexanderSwan]

@@davul01 sorry, I’m honestly not following your description/question. Good luck on your analysis!

[@nikolalabudovic5926]

Hi. Is there a way to do mediation in Jasp or jamovi if I have a categorical or ordinal mediator? I can't find the answer anywhere.

[@AlexanderSwan]

I do not believe they can do this right now. Only continuous variables can be used in these models, because they are constrained in the R packages

[@nikolalabudovic5926]

@@AlexanderSwan I actually tried doing it in Jasp and it didn't work, but I was able to do it in Jamovi so yeah, it worked in the end. Thanks for the reply!

[@AlexanderSwan]

@@nikolalabudovic5926 I couldn't remember which I had messed with recently! Glad it worked!

[@emmav1092]

Can I also use this with two predictors? Thanks!

[@AlexanderSwan]

Not to my knowledge. But it depends on what you mean by predictors.

[@emmav1092]

@@AlexanderSwan I mean two independent variables, one mediator, one dependent variable. I just found out PROCESS is unable to do this, therefore I am searching for another method/program I can use.

[@elsaboulard744]

@@emmav1092 I think you just have to do two mediation models (one for each predictor)