Hi there,
I was thinking about Value at Market Risk measure which takes into account the correlation of risk factors from style analysis which is based on (multi-factor) regression analysis. One of the assumptions of regression analysis is that the correlation of independent variables is zero (the violation is referred to as multicollinearity) so I would think there would be an assumption that the correlation of risk factors is zero for Value at Market Risk, but this is not the case.
Can anyone please explain what I am not seeing and help bridge the gap of my understanding of regression analysis and the Value at Mark Risk concept?
Thanks!!
According to Lhabitant, he uses the CSFB/Tremont sub-indices as the factors for the multi-factor regression analysis; and the sub-indices are supposed to be uncorrelated to each other.
Thanks, but how does that tie to VaMR?
VaMR, by definition, is a VaR due to market/style index moves.
You are right sanfran girl, VaMR does not assume independence.
Multicollinearity concept is for theoritical models where it is useful to have independent style factors. However, it reality, you never have completely independent factors (i.e. all models exhibit multicollinearity)