Multilevel mediation in SPSS (video 1): Testing 1-1-1 models using Rockwood's MLmed macro

[@katiekim458]

Hi Mike, all your videos save my life!!! Thank you so much!!

[@Furberoo]

Hi Mike, thanks so much for this video! Are you using a Long File (e.g., Person-Period data set) for this analysis? I can't tell from the video.

[@박선웅-b4s]

Hi Mike, Thank you very much for the video. It's extremely helpful. I want to know more about the meaning of "Test of Indirect Contextual Effect(s): Between - Within" as Rockwood's manual doesn't contain any information about it. Can you explain it more or provide any reference?

[@zselykepap4386]

Hey Mike! Clear and focused tutorial as always. Thank you for all your work! I have a question regarding the 1-1-1 model you present here: Can we add a dichotomous mediator in this model? Would that need a different setup, or can MLmed automatically adapt estimation to the dichotomous predictor? (If it is possible at all...) Thank you!

[@Zwaehn100]

Hi Mike, Thank you for the great tutorial! I however seem to be getting the same error message every time I try to even run the program. Several lines of diverse names such as "Matrix - 'R_NC' is undefined", "Matrix - 'COV-Ran' is undefined", etc. Would you know why this is happening and how I can resolve this issue? I originally though it was due to the folder path name, though I have now removed all special cases, numbers and spaces and still get the same issue. Thank you very much in advance! :)

[@hufs05]

Hi Mike!! Thank you so much for this Video!! I have a question. Is it possible to make a Bootstrap with MLmed in case of a small sample size? Thank you!! :)

[@tmpcox]

Great video! I maybe have a 'stupid' question: how do you account for more covariates/control variables? Just use the syntax by coping and filling in the predictors?

[@vickymay3143]

Quick question - what does adding a covariate actually do? Does it the include the variable as another mediator, or another X term? I want to control for a specific variable, and I'm not sure whether to include it as a parallel mediator or as a covariate

[@yanchenzhang5623]

Dr. Crowson, thank you for the fabulous video! A Q about the contextual effect in the output. Does a significant contextual effect signify potential ecological fallacy? Thank you again! @Mike Crowson

[@cheekyxoxmonkey]

Quick q - Is the "z test" reported alongside the Montecarlo in the indirect output the Sobel test? Also, is the "Effect" a beta value when it is reported? Thanks in advance

[@dorialu20]

Hi, Mike, thank you very much for sharing. I'd like to ask that if I run that 1-1-1 model, can I present only individual-level effects on my paper? or should I present level1 and level2 effects?

[@vickymay3143]

I know this is a silly question, but what do you actually mean by within/between effects. If you were to report the mediation analysis, saying that psychological safety didn’t mediate the relationship between colleague support and a promotive voice, which would you use. I’m struggling to interpret the output. Please answer, I am trying to learn how to do this for my dissertation.

[@mikecrowson2462]

Hi Vicky, this is not a silly question at all. If you have clustering in your data - a major reason for performing multilevel analysis - then it is possible that part of the total mediated relationship involving X, M, and Y occurs within groups and part of it can occur between groups. Think about it as an extension of standard multilevel regression, where you might include a predictor at Level 1 that has been centered at the cluster/group mean and that mean being re-introduced as a predictor of the random intercepts at Level 2 to capture a contextual effect. If I leave X uncentered centered at the grand mean, then the regression slope at Level 1 may reflect not only within cluster/group association, but also between-group association if X varies not only within groups but also between groups. When you center X within cluster, then the within-groups (L1) regression coefficient is purged of the between-group information, giving you a clearer picture of the within-association. If I re-introduce the mean of X (from Level 1) at Level 2 as a compositional variable, it can be used to explain variation in the between-group differences in the group means for the outcome variable. In this way, we have two regression coefficients that capture part of the total association between X and Y - that part that lies at Level 1 (within-groups) and part the part that lies at Level 2 (between-groups). Logically then, we can decompose a mediated effect of X on Y (where X, M and Y have all been measured at Level 1) into mediation as it might occur within groups (level 1) and between groups (level 2). Now, if you are simply not interested in testing whether part of the total mediation occurs between groups and want to focus only on what is occurring at Level 1, then you can go with the model (where I talk about only within effects being tested, see second example in presentation). However, if you are interested in testing both within and between portions of the total mediated effect, then you can go with the first example in my presentation. Regarding which to use when reporting: I would say that it's not really an issue of which to use, but rather how much emphasis you place in reporting the effects at the different levels. If your research question pertains largely to mediation as it occurs at Level 1, then you can emphasize that when writing up. But if it is the case that you did find evidence of mediation at Level 2, I certainly would not choose to ignore that finding (even if most of your emphasis theoretically is still at Level 1). At any rate, that's my two cents (albeit a long-winded two-cents!). By the way, if you want to spend more on the within- and between-issue, I have a new video and Powerpoint (the latter goes into more detail on this topic) here: kzbin.info/www/bejne/rma9Zn6PhKmqhM0 . Good luck with your dissertation!

[@vyeniaras]

It gives the following error when I try to install: Error reading customer dialog specification. Any ideas?

[@gulafshan9388]

Thank you very much Mike. I was looking for some videos on MLmed macro from a long time. I would request you to please help us to interpret the results so that we can actually understand the results. Thanks again

[@mikecrowson2462]

Hi Gul, thanks for your comment. Be sure to download the Powerpoint underneath the video description (basically drive.google.com/file/d/1LzWc4dKnE_N_BvNhe6OlADW-MNWlcrfY/view). It goes into interpretation of indirect effects using confidence intervals. I also provide some figures in there to help you visualize what relationships are being examined. Cheers!

[@syedjamalshah5284]

Miss afshan how you did that? did you find any way to interpret it?

[@maxinemacdonald20]

Hi Mike, I realise you uploaded this more than a year ago but thought I'd comment in the hopes you'd see this and be able to maybe help me figure out if I should be using a 1-1-1 model and not a 2-1-1 model? I am looking to see if insight mediates the relationship between symptoms and distress. However, my data is clustered as individuals have completed the same measures multiple times. Would this be the 1-1-1 model? Thanks so much if you see this!

[@NofarDuani]

Hi Mike, Thanks for making this video. I am trying to follow your instructions but no matter what I do I keep getting an error message that says: Invalid file specification, and that the file can't be accessed. Any idea what might be happening?

[@brookeedelman6416]

Hi! Is it possible to do this model with all binary variables? My IV, mediator and DV are all binary.

[@saleemazhar8100]

Hi Mike, i tried many times but Macro can not be installed. Please explain clearly ho to install. Both file gives error.

[@ergunkara1867]

hi Mike I read Nicholas Rockwood thesis figure out to do analysis so that video is very usefull video. but this is still beta version right?

[@mikecrowson2462]

Hi there. As far as I can tell it's still beta version. Thanks for checking out my video! Best wishes!

[@gulafshan9388]

And can you let us know if moderated mediation is possible in HLM?

[@mikecrowson2462]

I believe the macro is set up to be able to do moderated mediation. That's still something I have yet to explore with MLmed. But I'll put it on my 'to-do' list. best wishes.

[@gulafshan9388]

@@mikecrowson2462 thanks Mike. I have actually done moderated mediation in MLmed but am facing difficulty in interpretation. That's why would be greatful to you if you make such video or guide me. Regards

[@hinashahid3087]

@@gulafshan9388 can you please share how did you run moderated mediation using MLMED?

[@kevindurham]

Great mediation video, I feel so relaxed that I'm going to sell all my weed, thank you, so much, let's be KZbin pals?