Options

Every setting of a run is a field of one option tree. hetlearn.defaults() returns the full tree with its default values; Options reference lists every field. The pages in this section take one group at a time and show working code.

How options reach a run

hetlearn.defaults()  <  config/master_options.m  <  the scenario's cfg

hetlearn.load_options merges the three field by field (later wins) and checks every name and value. A misspelt field stops the run with hetlearn:options:unknown and the full option path; a removed option stops it with hetlearn:options:removed and names the replacement. The merged tree is stored with every result as res.cfg.

Three ways to set an option:

% 1. For the whole project: config/master_options.m
o.switching.intensity = 2;

% 2. For a scenario and everything that inherits from it: in its file
function cfg = my_scenario()
cfg = a1_learn();
cfg.id = 'my_scenario';
cfg.switching.intensity = 1;
end

% 3. For one run, in memory
cfg = a1_learn(); cfg.id = 'try_chi1'; cfg.switching.intensity = 1;
res = hetlearn.run(cfg, 'save', false);

Some groups are replaced as a whole rather than merged: conditions, policy.loss, expectations.types, expectations.transforms, shares.init, shares.fixed. Free-form subtrees are merged without name checks: beliefs.init.values, beliefs.plm.by_var, beliefs.plm.derived, gain.by_var, shares.fixed_n2 (as a struct), shocks.add, report.

To see the options a scenario ends up with:

opts = hetlearn.load_options(a1_learn());
opts.gain

The pages

Page	Options and functions
Expectations	expectations.mode, expectations.variables
RE agents	expectations.re_variant, expectations.re_info
Timing and information	beliefs.info, expectations.re_info, history.re_states, history.realtime
Perceived laws of motion	beliefs.plm.default, .by_var, .derived, .var
Gains	gain.rule, gain.value, gain.by_var, gain.koz.*
Initial beliefs	beliefs.init.*, beliefs.update, history.learn
Shares and switching	shares.*, switching.*
Projection facility	projection.facility
History modes	history.mode, history.inversion.*, history.kalman.init, history.learn
Forecasts, shocks and conditions	forecast.*, shocks.*, conditions, compare_with, engine.*
Monetary policy	policy.*
Density forecasts	density.*
Shock decompositions	hetlearn.decompose
Forecast evaluation	hetlearn.evaluate, hetlearn.evaluate_tables
Further expectation rules	expectations.types, expectations.transforms, shares.init, shares.fixed; seven rules from the literature

Defaults and the project baseline

The examples use the SW07 scenarios of the project, which inherit the project baseline of config/master_options.m (KOZ gain from 0.05, \(\chi = 2\), memory 0.8, t-1 information for both types, filtered history, facility 'alm'). Where a page quotes a default, it is the value in hetlearn.defaults(); the baseline value is given next to it when they differ.