Hi,
In my study I have to test a mediation hypothesis in a multilevel (2-1-1) non-longitudinal model, with a categorical IV. I have access to R and SPSS, eventually to MPlus as well. As I initially plan to analyze the data using multilevel modelling in R (lmer/blmer), we decided for N=150.
A bit about the design: participants are randomly assigned to one of three experimental conditions (between subject, level 2, categorical), and then evaluate 4 profiles (within-subject ourcome, level 1, numerical). The mediator is a measure of stereotypes (numerical, level 1). I am setting my model on the base of the profile ID, as my main hypothesis is the effect of the experimental conditions on the different profiles (I am not interested in individual differences).
I have considered Bootstrapping in R (by creating a code myself), or the Macro for Monte Carlo Mediation in SPSS, but I am not sure how to approach these methods given my specific study. Also, I know about MSEM in MPlus, but I think I would need a much bigger sample and I should find a way to get access to MPlus (we are still collecting data so I can consider changing things around this issue).
Thank you for you help!
Hi Martina,
here you can find a great tutorial to conduct multilevel mediation in R using lmer-function: https://cran.r-project.org/web/packages/mediation/vignettes/mediation.pdf
If you have the possibility to use Mplus, your sample size may be sufficient. Just use observed cluster means (i.e., Level 2 means) if the estimation with MSEM does not work.
Best,
Kai
Hi Kai N. Klasmeier ,
Thank you very much for your input! I had seen the package, but I initially thought it would not apply to my study. With this article I can finally see that I can actually use it. Great also that MPLUS is an option with that clustering that you suggest.
Best,
Martina
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS