Background
The current bond dataset contains monthly_return and monthly_return_winsorized, but the two fields are presently identical and the latter has not actually been winsorized. The current schema therefore gives a misleading semantic signal and does not yet contain the approved bond excess_return field.
A refreshed dataset is expected to extend coverage through December 2025 and add the excess return defined in #8. This Issue tracks validation and ingestion of that delivery separately from portfolio construction.
Required validation
- Confirm the maximum observation month is December 2025 and identify any coverage gaps.
- Compare
monthly_return and monthly_return_winsorized observation by observation.
- Either apply and document a real winsorization rule, rename/remove the misleading field, or prohibit it from analytical use.
- Confirm
excess_return is computed as monthly_return minus the aligned one-month U.S. T-Bill return.
- Validate field units, missingness, duplicate bond-month keys, extreme values, and identifier stability.
Acceptance criteria
Related issues
Background
The current bond dataset contains
monthly_returnandmonthly_return_winsorized, but the two fields are presently identical and the latter has not actually been winsorized. The current schema therefore gives a misleading semantic signal and does not yet contain the approved bondexcess_returnfield.A refreshed dataset is expected to extend coverage through December 2025 and add the excess return defined in #8. This Issue tracks validation and ingestion of that delivery separately from portfolio construction.
Required validation
monthly_returnandmonthly_return_winsorizedobservation by observation.excess_returnis computed asmonthly_returnminus the aligned one-month U.S. T-Bill return.Acceptance criteria
monthly_return_winsorizedno longer claims a transformation that was not performed.excess_returnis present and matches hand-checked formula calculations.Related issues
monthly_returnconfiguration.