Dear Tor and co-developers of Mediation Toolbox,
I am contacting you to ask a few questions rather than raising an issue with the toolbox :) I am using the mediation toolbox for analysing a longitudinal (2 sessions - 3 groups) multimodal (functional, structural, DTI) brain imaging+behavioural data. The data has been preprocessed, therefore, we are looking into using the toolbox for not the raw data (as per examples provided with the toolbox) but the preprocessed one. We have multiple potential mediators (large size data - 67 patients x 87 values or features) and I am concatenating the matrices into one large variable (e.g., M = [Vol1 Vol2 fc1 fc2 sc1 sc2 ...], 67 x N). However, I have a few questions:
- The toolbox seems to be able to deal with categorical/multinomial (either binomial or more) variables as the independent variable (X - intervention/patient groups) and also as covariates (sex, centre, etc.). However, we seem to have issues/errors with this... of course, the two variables for M and Y have been standardised.
- We are very interested in finding out which columns of the matrix M (mediator as explained above) are chosen as potential mediators and how good of mediators they are individually. However, we just have two variables, being paths and stats, as outputs. Is there anyway of finding out how the features/columns of matrix M perform as mediators and which ones are the best ones?
- Unfortunately, we do not seem to be able to plot the outputs either. This will help with better understanding the underlying elements of the mediation process.
At the moment, the piece of code that we are using is as follows:
[paths, stats] = mediation(X, Y, M, 'boot', 'plot', 'verbose', 'covs', covariates, 'covnames', covnames, 'bootsamples', 10000);
Any suggestions on how we can improve our analysis or find out more about the details of the mediation would be very much appreciated.
Warmest wishes and please stay safe,
Amir Dehsarvi
Research Fellow | Aberdeen Biomedical Imaging Centre | School of Medicine, Medical Sciences, and Nutrition | University of Aberdeen