iFit/max_likelihood

 c=max_likelihood(Signal, Error, Model)

 Maximum-likelihood criteria, which minimizes the Normal log-likelihood
 the return value is a vector, and most optimizers use its sum (except LM).
 (|Signal-Model|/sigma).^2 + 1/2 log(2pi sigma)

 A good fit corresponds with a criteria lower or equal to 1.

 <http://en.wikipedia.org/wiki/Maximum_likelihood>
 (c) E.Farhi, ILL. License: EUPL.