FRED-SD Predictor Scope
Current group: FRED-SD predictor scope
This conditional group is shown only when the selected FRED source panel includes FRED-SD. It restricts which state-level source columns can become candidate predictors x and records the native-frequency evidence needed by Layer 2.
These axes do not choose a mixed-frequency model. They close the source selection contract: which states, which FRED-SD workbook series, and how strictly the selected set must agree on native frequency.
If the source mode is custom-only, or if the FRED panel is fred_md /
fred_qd without FRED-SD, these axes are hidden by default. Imported recipes
with stale non-default FRED-SD choices keep the axes visible as incompatible
choices so the user can remove or revise them.
Axis |
Choices |
Default / rule |
|---|---|---|
|
|
Default |
|
Census regions/divisions, |
State Scope. Default |
|
|
State List. Default |
|
|
Series Scope. Default |
|
|
Series List. Default |
State Scope vs State List:
fred_sd_state_groupis the user-friendly group choice. It can select all states, census regions/divisions, the contiguous-state set, or a custom group.state_selectionis the lower explicit-list switch. It exists so custom state groups and recipe imports can say whetherleaf_config.sd_statesmust be read.
Series Scope vs Series List:
fred_sd_variable_groupis the user-friendly workbook-series group choice. It can select all FRED-SD series, predefined economic groups, predefined t-code-review groups, or a custom group.sd_variable_selectionis the lower explicit-list switch. It exists so custom series groups and recipe imports can say whetherleaf_config.sd_variablesmust be read.
Layer 1 output:
selected states;
selected workbook series;
source sheets and series metadata;
native-frequency report for selected FRED-SD series.
Layer 2 boundary:
fred_sd_mixed_frequency_representationchooses calendar alignment, dropping policies, native-frequency block payloads, or mixed-frequency model adapters after Layer 1 has loaded and reported the source columns.MIDAS or other model-side mixed-frequency behavior is Layer 3 training logic.
Selector YAML:
path:
1_data_task:
fixed_axes:
dataset: fred_sd
frequency: monthly
target_structure: single_target
fred_sd_state_group: custom_state_group
fred_sd_variable_group: custom_sd_variable_group
state_selection: selected_states
sd_variable_selection: selected_sd_variables
leaf_config:
target: UR_CA
horizons: [1]
sd_states: [CA, NY, TX]
sd_variables: [UR, PAYEMS]