Mediation analysis with categorial IV and Mediator and continuous DV

[Marissa4usa]

Hi all,
I've been trying to figure out the specifics for a mediation analysis where both the IV (4 levels) and the mediator (11 levels) are categorical with a continuous DV.
I do know that it is a 3- step process. I ran an ANOVA for the relationship between IV and DV and a chi-square analysis for the relationship between IV and mediator.

But - I am a little confused about how to set up the mediation analysis in SPSS. Will I need to dummy code the IV, both the IV and mediator or just the mediator? Would I run this analysis as a linear regression or logistic? Can I even run this analysis in SPSS?

I'm really not trying to have other do my work. I'm just genuinely confused as to what steps I need to take for this analysis.

Thanks!

---

Members don't see this ad.

 

[CheetahGirl]

I'm not going to be able to answer your question without wrapping my brain around your questions. Frankly, I don't have the energy to spare right now. However, if you put into Google "mediation model in SPSS" you get some articles that may be helpful to figure out how to run your analyses.

Also, if you are in psych program, one of your Stats TAs may be better equipped to help you out.

Mediation models are the most fun to work with, so hopefully you'll be able to figure this out without too much agony.

Good luck!

 

[cara susanna]

I wish you had asked that question before I took comps! They asked that very same question.

 

[Ollie123]

Not sure I've ever seen that done before.

Is it TRULY categorical or more of an ordinal variable? I've seen some suggestions that you can still use the standard bootstrapping macros on ordinal data, but that obviously won't work for true categories. I suppose you could dummy-code it and then run the multiple mediator analyses....I'd need to REALLY look into it before saying that would definitely be okay to do since it might be a gross violation of assumptions and I'm not even sure how well they handle dichotomous mediators.

Believe it or not, there is a facebook page for statistical mediation that may be a good place to ask. While I'm sure Preacher, Mackinnon, etc. get obscene amounts of email about things like this if you truly do have a unique situation and can't find any guidance in the literature on this they might be generally interested and willing to offer suggestions, or at least reply to a brief email inquiry.

Let us know if you come across an answer. I have a feeling it won't be an easy answer no matter what it is. You could certainly test all possible combinations, but interpreting something like that would prove a nightmare.

Cara - That is a HORRIFIC comps question. I consider myself to be at least reasonably on top of my stats, and that is way out there even for me.

 

[cara susanna]

Well, it didn't outright ask it, but the question asked us to do a mediation analysis on two IVs that I couldn't operationalize in any other way. I just made the discrete variable continuous by measuring it differently.

 

[Neuropsych2be]

Well, it didn't outright ask it, but the question asked us to do a mediation analysis on two IVs that I couldn't operationalize in any other way. I just made the discrete variable continuous by measuring it differently.

Brilliant!

 

[cara susanna]

Why thank you! I hope my graders agree with you

 

[Marissa4usa]

I'm not going to be able to answer your question without wrapping my brain around your questions. Frankly, I don't have the energy to spare right now. However, if you put into Google "mediation model in SPSS" you get some articles that may be helpful to figure out how to run your analyses.

I've done just that, and virtually every single combination of continuous and categorial variables comes except my combination.

Mediation models are the most fun to work with

Debatable

Is it TRULY categorical or more of an ordinal variable?

Definitely not ordinal

Believe it or not, there is a facebook page for statistical mediation that may be a good place to ask. While I'm sure Preacher, Mackinnon, etc. get obscene amounts of email about things like this if you truly do have a unique situation and can't find any guidance in the literature on this they might be generally interested and willing to offer suggestions, or at least reply to a brief email inquiry.

Yeah, I know about that one. I posted my question there already. I hope somebody can help me...otherwise I'll be in trouble...or I'll just have to come up with some lame replacement analyses.

 

[CheetahGirl]

Let us know if you come across an answer.

Agreed! Marissa4usa, let us know what you end up doing...

 

[sydb1367]

Well, it didn't outright ask it, but the question asked us to do a mediation analysis on two IVs that I couldn't operationalize in any other way. I just made the discrete variable continuous by measuring it differently.

Way to Kobayashi Maru your comps, cap'n Kirk. that's awesome