What to report in a mediation analysis

#1
Hello
Given my comprehensive lack of statistical knowledge, I have gained a lot of information over the last five months browsing this forum.

But now I'm stuck and would dearly appreciate any assistance.

When reporting a mediation analyis for an APA style write up -should I include the R2, F and t-stat as well as the p value - or is the convention only for the F-stat and p value?

Standardised coefficients have been placed on a Figure.

Again, any assistance would be greatly appreciated - there is a notable parcity of info regarding this, out "there", so any feedback would be great.

And whilst the relationship has been parasitical thus far, I hope (pray, sweat, swear, weep) one day I may be of some assistance to someone here!

With thanks
Honey
 
#2
Dear Honey

Welcome to the community, and hope I can soon learn a lot from you :)

I don't know about the APA style, and about mediation analysis! But based on your post, it seems something like regression. If so, I can say those important parts for the model which should be reported are the F statistic, P, and adjusted R-squared. It is also good to report the difference in the adjusted R-squared values for each model compared with its previous model, if you have more than one model. Well the P and standardized coefficient are important for predictors.


Hi Honey

So my table - if I have the adjusted R2, (as they were run as three seperate regressions, only one model per analysis, so no differences between adjusted R2 to report) the F stat and the SE of the beta weights - what would I call the table??? It has coefficients/ANOVA etc. Any hints here??? Maybe I should just call it Barry!
Lol, you can call it "the results of the regression analyses", or any name that can include all the analyses reported at the same time. But that syndicated table would be a good one IMHO. Also would be good to report the P values for the predictors, because I think you are only reporting their beta values without corresponding P values.

Thanks again Victor - very kind of you to help
As for me helping - ooohh don't hold your breath. I'm good for Wiggles lyrics and hints on getting weetbix of curtains, and thats about it!
lol :) You are so welcome, that would be good too :p
 
#3
Hi Victorxstc - thanks so much for your reply - you be fast!!!!!! I be slow.

Yes- it is a regression - basically a mediator varaible is hypothesised as being either a partial or full 'cause" of an IV's effect on the DV - so the IV is signifcantly related to the DV, the IV is significantly related to the Mediator, the mediator is significantly related to the DV after controlling for IV's effect on DV and the relationship between IV and DV becomes either non-sig (full mediation) or smaller (partial mediation) after controlling for the mediator. Well thats according to Baron & Kenny (1986) anyway. Actually sort of like a more involved ANCOVA??

So I have a figure which has the standardised beta's on it, my results section has the t-stat, direction and p value

So my table - if I have the adjusted R2, (as they were run as three seperate regressions, only one model per analysis, so no differences between adjusted R2 to report) the F stat and the SE of the beta weights - what would I call the table??? It has coefficients/ANOVA etc. Any hints here??? Maybe I should just call it Barry!
Thanks again Victor - very kind of you to help
As for me helping - ooohh don't hold your breath. I'm good for Wiggles lyrics and hints on getting weetbix of curtains, and thats about it!
 
Last edited: