Selon David Muldowney <
[hidden email]>:
> Hi
> I'm very new to QuantLib so forgive my innocence here.
> My problem is a regression calculation, ex-post Alpha and Beta given a set of
> returns for a fund and a benchmark.
> Calculating Beta and Alpha boils down to getting the best-fit line (min.
> least squares problem I guess) and then reading it's slope and intercept.
> I've looked at the documentation and I think my best bet seems to involve
> LineSearch (in the Optimisation namespace), but I'm open to correction.
> I'm doing the code in VB.Net but if anyone can give me an example of how this
> class (or ones more appropriate to my problem) are used (C++ / C# would do) I
> would be immensely grateful
> Regards
> David
>
Hi David,
The optimization framework has been made for non linear problems. It should
allow you to treat linear ones (*) but it is not optimal. I think that a
regression problem reduces to a symetric matrix inversion that can be solve
using direct or iterative methods.
Nicolas
(*) Indeed, you have to define a cost function class (deriving from CostFunction
class) describing your problem (ie implementing the value(const Array& params)
method). At the end, you have to use the Simplex method to solve your problem.
--
Nicolas Di Césaré
http://nicolas.dicesare.free.fr
Locked
