Data Layer#

Layer 1 defines the forecasting dataset: source, target, predictor universe, geography, sample window, horizons, and regimes.

Core source axes:

Axis	Common values
`custom_source_policy`	`official_only`, `custom_panel_only`, `official_plus_custom`
`dataset`	`fred_md`, `fred_qd`, `fred_sd`, `fred_md+fred_sd`, `fred_qd+fred_sd`
`frequency`	`monthly`, `quarterly`
`information_set_type`	`final_revised_data`, `pseudo_oos_on_revised_data`
`vintage_policy`	`current_vintage`; `real_time_alfred` is schema-present but future.

Target and horizon axes:

Axis	Common values
`target_structure`	`single_target`, `multi_target`
`variable_universe`	`all_variables`, `core_variables`, `category_variables`, `target_specific_variables`, `explicit_variable_list`
`horizon_set`	`single`, `custom_list`, `range_up_to_h`, `standard_md`, `standard_qd`
`sample_start_rule`	`earliest_available`, `fixed_date`, `max_balanced`
`sample_end_rule`	`latest_available`, `fixed_date`

Use leaf_config.target for single_target, leaf_config.targets for multi_target, leaf_config.target_horizons for single or custom_list, and leaf_config.max_horizon for range_up_to_h.

FRED-SD geography is controlled by target_geography_scope, predictor_geography_scope, fred_sd_state_group, state_selection, fred_sd_variable_group, and sd_variable_selection.