R - Mediation Analyses with the multilevel package

[@jithesha3214]

Dear Erin, Great videos. I am looking for resources on how to do a mediation analysis with a panel/longitudinal data. Can you please advice?

[@StatisticsofDOOM]

Unfortunately, I don't have any good resources for those ... mostly because I haven't don't panel type analyses, mostly I stick to multilevel models when I have longitudinal data.

[@arghavanomidi3718]

Thanks for this video! any package/function suggestions for multilevel mediation with categorical data?

[@StatisticsofDOOM]

Note that multilevel here does mediation, not multilevel models (well it might but I would recommend lme4 or nlme). I don't know which variable you mean is categorical but nlme and lme4 can handle most of it.

[@miathomaidou4090]

Thanks so much for the video! I have a question regarding my data. I have a dichotomous IV, continuous M, and categorical (ordered) DV. Would I be able to run this mediation as shown in this video? Is there a need to separately dummy code the DV? MANY thanks!

[@StatisticsofDOOM]

You should consider a logistic regression for a categorical DV, this version is assumes you have a continuous DV.

[@miathomaidou4090]

@@StatisticsofDOOM thanks very much for your response! I have done a logistic regression and I wanted to see next whether there was a mediation. This example didn’t work indeed, still really helpful for me to understand why it doesn’t work. I have now done a mediation by using an sem function for an ordered outcome, and this seems to work and makes sense with my data. PS Your videos have really changed my work for the better, thanks very much!

[@StatisticsofDOOM]

@@miathomaidou4090 SEM is also a good choice (it's the same math for ordinal regression, so go with that!).

[@xcathyxMacx]

Hi Erin, Thanks for the great video. Do you happen to use the mlma package at all? If so, do you have any advice on good resources for this?

[@StatisticsofDOOM]

I do not use it - best suggestion is the CRAN documentation: cran.r-project.org/web/packages/mlma/mlma.pdf

[@xcathyxMacx]

@@StatisticsofDOOM Thanks!

[@jessemaombi]

My X variable is a factor with 3 levels... My Y variable is a continous variable while my M variable is a factor with 8 levels... How do i test the significance of the mediation effect because ive tried both the mediation package and the multilevel package...using the sobel test, everything is okay except when it comes to refressing X on M... The standard values, t values and p values are NA.. The indirect effect is given but the SE and Z. Value are NA... Please help me.. Am doing a school project

[@StatisticsofDOOM]

Likely because the X and M variables are factors, which makes for lots of indirect effects. Are there enough people in each of these combinations?

[@huanlee9325]

Dear Erin, Thanks for the video,it is help me a lot. I have a big question, could you help me? i must run a multilevel mediation model and then i use lavaan, but when i want to check mediation analysis with bootstrap， i can not find the method by R. hope you help me figure it out，thanks very much!

[@StatisticsofDOOM]

Lavaan has bootstrapping functions built in - and mediation as well - check out lavaan.ugent.be/tutorial/mediation.html

[@anamatic952]

Dear Erin, I've looked at this and the other video on mediation but kind of stuck on what to do when I have multiple (parallel) mediators (more than two). do I just simply do each of the steps for each one of the mediators or might I hope for maybe one more video on this matter? thanks!

[@StatisticsofDOOM]

Yeah, I am not sure if I would know how to three either - you would have to figure out how to calculate the indirect paths for each one ... you might consider a structural equation model if you have a more complex picture.

[@a.n.4066]

Great Video. I'm searching for a Mediation Analyses with Multilevel with hierarchical structur.

[@StatisticsofDOOM]

Not sure I understand the question. This video covers MLM and mediation.

[@oui0384]

@@StatisticsofDOOM this video seems to be a regular mediation not MLM

[@StatisticsofDOOM]

@@oui0384 I noted in the description the name of the package is multilevel, and added an update to clarify this is not a multilevel model (I do have those examples on the channel though).