1
$\begingroup$

I am new to the mediation analysis and i am trying to find my way...

So, I have the following dataset:

X's : 5 independent variables (D, E, F, G, H), where H is the sum of the other 4 (probably not very wise to be in the analysis, and also the models produces an error whrn i include it, and so i left it out ), continuous

M's : 1 mediator (B), continuous

Y : 1 response (I), continuous

Z : 1 covariate (not in the output yet), continuous

I also have some missing values for some of them (around 6-7%).

So, i have first to use MICE to input and then run the mediation analysis. But, as i found out it is not straightforward to combine them, so i used this example here. They use the lavaan package in R.

The output i got, is the following :

enter image description here

Can somebody please help me interpret this output here ? I am very confused with the several predictors in the model and i don't know how to assess mediation or not...And even if it is appropriate to add all these variables in a model like this..

And finally, if you know something more elegant about the combination of MICE and mediation analysis...

Thanks

Improve this question
$\endgroup$
5
  • $\begingroup$ You can try looking into mediation package or medflex package in R. Mediation package is more mature and has more features but has some limits for model type. (Feel free to make this a comment once you see this). $\endgroup$
    – V. Aslanyan
    Commented Jul 31, 2018 at 20:45
  • $\begingroup$ Thanks @V.Aslanyan ! I was indeed looking at the mediation package..But i am not sure how to apply the 4 predictors, because it "uses only the first" as the warning says...And also, i tried to use mediations function which is precisely for many datasets (as he case of multiple imputation), but i get an error and i can't figure it out... {Error in rep(1, nrow(dataarg)) : invalid 'times' argument} $\endgroup$
    – GiannisZ
    Commented Aug 1, 2018 at 8:52
  • $\begingroup$ I am not familiar with any literature concerning multiple independent variables as treatment variable. As for the error, look what it is taking as your dataarg, seems like it does not have rows, maybe it requires your dataset to be in a different format. $\endgroup$
    – V. Aslanyan
    Commented Aug 1, 2018 at 16:56
  • $\begingroup$ Thanks again! Unfortunately there is not an example somewhere with mice() + mediation() ... And in their documentation they only mention it without an example! Anyway, thanks for you answer! $\endgroup$
    – GiannisZ
    Commented Aug 2, 2018 at 7:55
  • $\begingroup$ Don't know how much this will help you, but at this stage of our research we are looking into developing multiple treatment models and I found this paper. Please let me know if you find anything relevant too. I would greatly appreciate it. $\endgroup$
    – V. Aslanyan
    Commented Aug 24, 2018 at 19:09

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.