ACES LQ Implementation Interface Requirements

User options when invoking and running ACES Linear Quadratic approximation

Those user optinos need to be provided as a part of ACES LQ integration and its interface with Dynare. When not specified as interactive it is expected to be provided as a part of the model file model specification.

Option

Meaning and description

Poss./Def. val (in bold)

Ineractive [1]

planner_discount

as currently used in Ramsey

0.1-0.99

instruments

as currently used in Ramsey

use_aces_lq_solver

option to invoke ACES LQ unless separate application used

aceslq_seedvar

Choice of variable(s) that characterises the initial guess of the steady state of the state – used in the MATLAB steady state solver (may differ from the choice of the instrument)

User choice

aceslq_lambda_w

search Relaxation weight – used in the main work of relaxation method (MATLAB_LQ)

0.5

M

aceslq_maxiter

Maximum number of relaxation iteration

500

M

aceslq_accuracy

Accuracy used to compare real numbers (ACES)

A small real number e.g. 1E-6

aceslq_ubound

Specifies the upper bounds in the NAG search (ACES)

User’s choice as an option if ibound=0

aceslq_lbound

Specifies the lower bounds in the NAG search (ACES)

User’s choice as an option if ibound=0

aceslq_igrid

Used to ask for a grid search in SIM (ACES) – SETGD subroutine required if igrid=0

1 – grid search not required but user must provide the exact values of coefficients used in the simple rule

F&M

aceslq_ITMAX_OPT

max no of iteration in OPT (ACES)

20

aceslq_ITMAX_TCT

max no of iteration in TCT (ACES)

5000

aceslq_ITMAX_SIM

max no of iteration in SIM (ACES)

1000

aceslq_SIM_search_step_size

optional initial_SIM search step_size

0.5

F&M

aceslq_min_SIM_search_step_size

optional minimal SIM search step_size

0.01

F&M

Choose Optimiser

Choose Optimiser or exit

M

Expected for later releases:

aceslq_MCMC_model

model to be used in Optimal Simple Robust Rules function:i.e. original estimation model name for the MCMC draws

<model_name>

aceslq_MCMC_selection_method

Choose selection method: random

random or fixed

aceslq_MCMC_draw_interva

for Fixed: what draws to select interval

10

aceslq_MCMC_draw_number

how many

5000

aceslq_MCMC_start_end

and from which end: options

start or end

Notes [1]: M-Matlab only, F&M- Fortan and Matlab

DynareWiki: AcesLqImplementationRequirements (last edited 2010-09-17 13:46:50 by GeorgePerendia)