
basically i have this objective: i want to quantitavely narrow down the number of combination of possible explanatory variables that i have to use to discover the target fund composition. i will then use economic theory to come up with a sensible result
problem: since weights are set to be positive we don't use regressions but numerical optimization (the Tracking Error volatility is the target to be minimized). but i want to check for collinearity first to exclude a good amount of combinations. fact is this is not a proper regression and i am also aware that correlation is different from causality so i might accidentally eliminate the wrong input. still, i need to cut to size my workload.
i tried to read some papers and i found out some stuff on bayesian methods but they are far too advanced for the homework i am required to perform.
any ideas on how to proceed? 




For starters maybe review the phorum guidelines and learn to write like a grownup. 

