I have a correlation matrix that I produced in Excel using the monthly returns from 200 different trading systems over a X year period. So, the correlation matrix is 200 rows by 200 columns. How do I now go about determining the best combination of 10 systems that would give me the lowest correlated subset (i.e., lowest potential of aggregate sum of correlation values in a 10x10 correlation matrix)? Also, can this be done in Excel? Thanks.
2
-
$\begingroup$ Some form of factor analysis, maybe a singular value decomposition could be a start? $\endgroup$– Mats LindAug 12, 2016 at 16:09
-
1$\begingroup$ Mabye this is useful: quant.stackexchange.com/questions/2263/… $\endgroup$– Mats LindAug 12, 2016 at 16:58
Add a comment
|