I have the covariance matrix between log-returns of n variables. I suppose the distribution of the log-returns is normal for all the variables with average=0 but standard deviation in general $\neq$ 1. With the realized log return of n-1 variables I want to get an estimation of the last variable, let's call it K.

I thought about using Cholesky decomposition to get the triangular matrix L.

Then to get the best estimation I used:

$$\sum_{i=1}^{n-1} L_{n,i} * R_{i}$$ (where R is a vector of log return for n-1 variables)

And for the standard deviation I used:

$L{n,n} * std(K)$

I'm not sure this is right. In particular I suppose the general procedure is correct but there may be errors in "scaling". Maybe I have to rescale $R_{i}$ dividing them by the relative standard deviation to evaluate the best estimation and use $L{n,n}$ as the standard deviation of the best estimation.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.