Commit b685fe61 authored by Stéphane Adjemian's avatar Stéphane Adjemian
Browse files

Updated description of the optim option in the estimation command (mode_compute=1,3,7 and 4).

parent 05946cd6
......@@ -4401,6 +4401,7 @@ hessian (@code{hh}, only if @code{cova_compute=1}) in a file called
@file{@var{MODEL_FILENAME}_mode.mat}
@item mode_compute = @var{INTEGER} | @var{FUNCTION_NAME}
@anchor{mode_compute}
Specifies the optimizer for the mode computation:
@table @code
......@@ -4504,11 +4505,45 @@ parameters. Default: @code{1e-32}
Metropolis-Hastings simulations instead of starting from
scratch. Shouldn't be used together with @code{mh_recover}
@item optim = (@var{fmincon options})
Can be used to set options for @code{fmincon}, the optimizing function
of MATLAB Optimization toolbox. Use MATLAB's syntax for these
options. Default:
@code{('display','iter','LargeScale','off','MaxFunEvals',100000,'TolFun',1e-8,'TolX',1e-6)}
@item optim = (@var{NAME}, @var{VALUE}, ...)
A list of @var{NAME} and @var{VALUE} pairs. Can be used to set options for the optimization routines. The set of available options depends on the selected optimization routine (ie on the value of option @ref{mode_compute}):
@table @code
@item 1, 3, 7
Available options are given in the documentation of the MATLAB optimization toolbox or in Octave's documentation.
@item 4
Available options are:
@table @code
@item 'MaxIter'
Maximum number of iterations. Default: @code{1000}
@item 'NumgradAlgorithm'
Possible values are @code{2}, @code{3} and @code{5} respectively corresponding to the two, three and five points formula used to compute the gradient of the objective function (see @cite{Abramowitz and Stegun (1964)}). Values @code{13} and @code{15} are more experimental. If perturbations on the right and the left increase the value of the objective function (we minimize this function) then we force the corresponding element of the gradient to be zero. The idea is to temporarly reduce the size of the optimization problem. Default: @code{2}.
@item 'NumgradEpsilon'
Size of the perturbation used to compute numerically the gradient of the objective function. Default: @code{1e-6}
@item 'TolFun'
Stopping criteria. Default: @code{1e-7}
@item 'InitialInverseHessian'
Initial approximation for the inverse of the Hessian matrix of the posterior kernel (or likelihood). Obviously this approximation has to be a square, positive definite and symmetric matrix. Default: @code{'1e-4*eye(nx)'}, where @code{nx} is the number of parameters to be estimated.
@end table
@item
@end table
@customhead{Example 1}
To change the defaults of csminwel (@code{mode_compute=4}):
@code{estimation(..., mode_compute=4, optim=('NumgradAlgorithm',3,'TolFun',1e-5), ...);}
@item nodiagnostic
@anchor{nodiagnostic} Does not compute the convergence diagnostics for
......@@ -8719,6 +8754,9 @@ internals --test particle/local_state_iteration
@itemize
@item
Milton Abramowitz, Irene A. Stegun (1964): ``Handbook of Mathematical Functions'', Courier Dover Publications
@item
Aguiar, Mark and Gopinath, Gita (2004): ``Emerging Market Business
Cycles: The Cycle is the Trend,'' @i{NBER Working Paper}, 10734
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment