1

I only found a example in which the constraind on $\beta_i, \ i=1,...k$ would make them sum up to one as well introduces a constraind such that all $\beta_i$ are all greater or equal zero. Since I dont need a condition for the summation of the $\beta_i$ I'm not quite sure how to deal with this problem.

Since in this context a matrix $C$ of constraints is introduced such that $C\beta = c$, I'm not sure how to deal with $C\beta \geq 0$.

Any suggestions are very welcome!

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Druss2k
  • 783
  • 6
  • 22
  • I know I've to solve some kind of restricted minimisation problem such as $\psi = (y - X\beta)^T(y - X\beta) - 2\lambda^T(C\beta - c)$. What I dont know is how to introduce some kind of greater equal condition? My guess would be Kuhn-Tucker minimisation. – Druss2k Mar 13 '13 at 02:10
  • 1
    The solution will be found by looking into non negative least squares (NNLS). – Druss2k Mar 13 '13 at 02:22
  • 1
    You should post your comment as an answer (yes, you can answer your own question). – Glen_b Mar 16 '13 at 06:04

0 Answers0