Help System (web edition)

The iterative numerical algorithm needed for minimizing of the
residual sum of squares may be selected by the user by an extra
specification METHOD typed in the edit field. At the moment we have
the alternatives:
 METHOD=N     Newton-Raphson            (see e.g. Walsh, 1975 p.108)
 METHOD=M     Modified Newton           (see e.g. Gill & al., 1982 p.105-)
 METHOD=D     Davidon-Fletcher-Powell   (see e.g. Walsh, 1975 p.110)
 METHOD=d     As D above, but without 2nd derivatives (saves space)
 METHOD=H     Hooke-Jeeves              (see e.g. Walsh, 1975 p. 76)
 METHOD=G     Direct search in a grid of parameter space
              The search values for each parameter A must given as
              A=<lower_limit>,<upper_limit>,<step> 
              This last alternative exists mainly for searching for good
              initial values of parameters to be estimated.

In case of a linear model METHOD=N is default and in case of a non-
linear model METHOD=D is default.
References: Walsh: Methods of optimization, Wiley (1975)
            Gill,Murray,Wright: Practical Optimization, Academic Press (1982)
  S = more on extra specifications in ESTIMATE