With regard to the scoring its working perfectly apart from being a little slow importing the data, which is currently stored in csv files. The better way of doing this is to store and run the code directly from a database. So I am in the process of planning a move to postgres. Once processed and scored, R can then be used to interrogate the database and run analysis. It should be a lot quicker with a setup like this and also gives me the advantage of holding quite a large dataset of instruments and scores. With regard to actual analysis I have been exploring correlations/spread among similar asset classes, using an ACD framewrok. So far it looks good from the scope of being able to identify regime change. This is all being done with the plan of identifying the best times to be trading spreads etc.
Both. I'm frustrated that GARCH on R apparently does not really make it easy to use multi-variable fundamental models.
I never did anything with garch but there are some packages for multivariate garch (e.g. https://cran.r-project.org/web/packages/rmgarch/index.html). Did you try it or some other packages?
Have you tried Stackoverflow or Quantstackexchange? If you can create a reproducible example of the problem then maybe I can help you. There is also a function on this page that will allow you to create an example. As ingl posted R does have multivariate Garch packages ready to go mgarch is another one.