Blurb::
Determine how the final covariance matrix is computed 
Description::
\c covariance
(NL2SOL's \c covreq) specifies whether and how NL2SOL computes 
a final covariance matrix.

The desired covariance approximation:
\li 0 = default = none
\li 1 or -1 ==> \f$\sigma^2 H^{-1} J^T J H^{-1}\f$
\li 2 or -2 ==> \f$\sigma^2 H^{-1}\f$
\li 3 or -3 ==> \f$\sigma^2 (J^T J)^{-1}\f$
\li Negative values ==> estimate the final Hessian H by finite 
differences of function values only (using \c fd_hessian_step_size)
\li Positive values ==> differences of gradients (using 
\c fd_hessian_step_size)


Topics::
Examples::
Theory::
Faq::
See_Also::	
