Options reference

Every field of the option tree returned by hetlearn.defaults() (toolbox/+hetlearn/defaults.m). Values are listed default first. "Baseline" is the value the project sets in config/master_options.m where it differs from the default. scenarios/sw07/base/sw07_re.m lists every option as a commented line, a convenient template.

Precedence: hetlearn.defaults() < config/master_options.m < the scenario. Unknown names stop the run (hetlearn:options:unknown).

Identification

Option	Default	Values	Meaning	Page
id	''	text	Results folder results/<id>/; required	Forecasts
title	''	text	Title for summaries and reports
description	''	text	Description for summaries and reports
model	''	text	Model package under models/, or a folder path; required	Your own model
model_params	''	'' or a function name	Parameter file of the package; '' = spec.params. Baseline 'sw07_params_original'	Forecasts

Forecast

Option	Default	Values	Meaning	Page
forecast.origin	[]	a date (required)	Last quarter of data used	Forecasts
forecast.horizon	16	positive integer	Forecast quarters

Expectations

Option	Default	Values	Meaning	Page
expectations.mode	'rational'	'rational', 'learning', 'fixed_shares', 'switching'	Who forecasts how	Expectations
expectations.variables	{}	cellstr or 'all'	Variables with heterogeneous expectations	Expectations
expectations.re_variant	'anchored'	'anchored', 'model_consistent'	RE agents' forecast	RE agents
expectations.re_info	'current'	'current', 'lagged'	RE agents' information. Baseline 'lagged'	RE agents
expectations.types	empty	struct array or cell of structs with name, kind ('re', 'plm', 'rule', 're_mean'), rule	Population of forecasting types; replaced as a whole	Further rules
expectations.transforms	empty	structs with kind ('multiplier', 'diagnostic', 'myopia_anchoring'), types, scope, exclude, value	Transforms of the expectations; replaced as a whole	Further rules

Beliefs

Option	Default	Values	Meaning	Page
beliefs.update	'rls'	'rls', 'fixed'	Learn, or hold the initial beliefs	Initial beliefs
beliefs.info	'lagged'	'lagged', 'current'	Learners' information	Timing
beliefs.plm.default	'ar1'	'ar1', 'ar2', 'ar1_noconst', 'var1'	Default PLM	PLMs
beliefs.plm.by_var.<x>	none	shorthand or struct('regressors', {...}, 'constant', true)	PLM of one variable (free-form)	PLMs
beliefs.plm.derived.<name>	none	text	Derived regressor (free-form)	PLMs
beliefs.plm.var.variables	{}	cellstr or 'learning'	VAR members	PLMs
beliefs.plm.var.lags	1	positive integer	VAR lags
beliefs.plm.var.constant	true	logical	VAR constant
beliefs.init.method	're_moments'	're_moments', 'training_ols', 'user'	Initial beliefs	Initial beliefs
beliefs.init.date	[]	a date	[] = first filter date + 4 quarters
beliefs.init.values.<x>	none	.alpha, .beta or .phi, .R	Your values (free-form)
beliefs.init.training_range	[]	[start, end]	Required for 'training_ols'

Gain

Option	Default	Values	Meaning	Page
gain.rule	'constant'	'constant', 'koz'	Gain rule. Baseline 'koz'	Gains
gain.value	0.0585	(0, 1]	Constant gain; start of 'koz'. Baseline 0.05
gain.by_var.<x>	none	number	Gain of one variable (free-form)
gain.koz.mu	0.001	>= 0	Step size of the gain process
gain.koz.gamma0	[]	[] or (0, 1]	Start gain; [] = gain.value
gain.koz.Rg0	0	finite	Start of \(R^g\)
gain.koz.bounds	[0.001 0.25]	[lower upper], 0 < lower <= upper <= 1	Projection interval
gain.koz.error	'rls'	'rls', 'forecast'	Error that drives the gain
gain.koz.error_scale	1	> 0	Factor on the error
gain.koz.structure	'common'	'common', 'per_variable'	One process or one per equation
gain.koz.master	''	name or cellstr	Driver of the common process

Shares and switching

Option	Default	Values	Meaning	Page
shares.structure	'per_variable'	'per_variable', 'universal'	One share per variable or one for all	Switching
shares.master	{}	name or cellstr	Drivers of the universal share
shares.init_n2	0.5	[0, 1]	Initial learners' share
shares.fixed_n2	[]	number or struct by variable	Learners' share in 'fixed_shares'
shares.init	[]	struct by type name	Initial shares over N types; replaced as a whole	Further rules
shares.fixed	[]	struct by type name	Fixed shares over N types; replaced as a whole	Further rules
switching.rule	'discrete_choice'	function name	Switching rule	Switching
switching.intensity	0.51	>= 0	Intensity of choice. Baseline 2
switching.memory	0.6	[0, 1)	Memory of the fitness. Baseline 0.8

Projection facility

Option	Default	Values	Meaning	Page
projection.facility	'plm_alm'	'plm_alm', 'alm', 'none'	Stabilising device. Baseline 'alm'	Projection facility

History

Option	Default	Values	Meaning	Page
history.learn	true	logical	Learn over the history	History modes
history.mode	're_smoother'	're_smoother', 'inversion', 'kalman'	Source of the history	History modes
history.re_states	'smoothed'	'smoothed', 'filtered'	're_smoother' only. Baseline 'filtered'	Timing
history.realtime	false	logical	RE filter on data up to the origin	Timing
history.inversion.start	[]	a date	First inversion quarter	History modes
history.inversion.shocks	{}	cellstr	Shocks backed out
history.kalman.init	'stochastic'	'stochastic', 'smoothed'	Start of the Kalman history

Shocks and conditions

Option	Default	Values	Meaning	Page
shocks.source	'none'	'none', 'smoothed'	Shocks over the horizon	Forecasts
shocks.range	[]	dates	Quarters with smoothed shocks
shocks.exclude	{}	cellstr	Shocks set to zero
shocks.add	{}	{{name, dates, values}, ...}	Shocks added on top (free-form)
conditions	empty	struct array: variable, shock, dates, values, type ('level', 'delta'), reference ('self' or an id)	Exogenise a variable, endogenise a shock; replaced as a whole	Forecasts

Density forecasts

Option	Default	Values	Meaning	Page
density.enabled	false	logical	Switch on	Density forecasts
density.shocks	{}	cellstr	Shocks drawn (required when enabled)
density.sigma	'calibrated'	'calibrated', 'smoothed'	Standard deviations
density.sigma_range	[]	[start, end]	Range for 'smoothed' and the bootstrap pool
density.method	'normal'	'normal', 'bootstrap'	Draws
density.ndraws	200	positive integer	Number of draws
density.seed	0	integer	Random seed
density.quantiles	[0.05 0.16 0.5 0.84 0.95]	vector in (0, 1)	Reported quantiles
density.range	[]	dates	Quarters with draws
density.keep_draws	false	logical	Store every draw
density.fast_path	true	logical	Linear fast path for RE runs
density.demean	true	logical	Demean the bootstrap pool

Engine

Option	Default	Values	Meaning	Page
engine.step	'fast'	'fast', 'iris'	One-quarter step	Forecasts
engine.conditions_weighting	'std'	'std', 'none'	Shock weighting with more shocks than targets

Monetary policy

Option	Default	Values	Meaning	Page
policy.mode	'rule'	'rule', 'ramsey'	Policy	Monetary policy
policy.type	'commitment'	'commitment', 'discretion'	Ramsey type
policy.loss	struct('pinf', 1, 'ygap', 0.5, 'r', 0.1, 'dr', 0)	weights by target name	Loss weights; replaced as a whole
policy.discount	[]	[] or (0, 1]	[] = spec.policy.discount
policy.loss_horizon	[]	[] or quarters	Loss horizon

Comparisons and reporting

Option	Default	Values	Meaning	Page
compare_with	{}	cellstr of ids	Writes compare_<id>.csv	Forecasts
report.variables	{}	cellstr	Reserved; read by no code

Option errors

Identifier	Cause
hetlearn:options:unknown	A name that is not in the defaults
hetlearn:options:type	Wrong type, e.g. a struct where a value is expected
hetlearn:options:value	A value outside its list or range
hetlearn:options:missing	'fixed_shares' without fixed shares, 'universal' without shares.master, 'training_ols' without a training range, density without shocks
hetlearn:options:unsupported	A combination that is not supported, e.g. Ramsey with re_info = 'lagged'
hetlearn:options:removed	An option name that the toolbox no longer has; the message names the replacement
hetlearn:options:types, hetlearn:options:transforms	An invalid expectations.types or expectations.transforms entry