Bump hosted SVAR estimation window 2023Q2 -> 2025Q1#86
Merged
Conversation
The hosted server reported two different samples for the same model: model_summary read the committed replication artifacts (1992Q1-2025Q1, refreshed in #85) while forecast_uk and get_model_status reported 1992Q1-2023Q2. That was not just a label — _SVAR_EST_END truncates the estimation frame in _estimate(), so the live forecast discarded seven quarters of data the package already ships. The constant is deliberately pinned so hosted results only move on a deliberate refresh; this is that refresh. Held one quarter short of the 2026Q1 data edge, matching the repo, so the forecast origin stays outside the estimation sample. Measured effect (seed 0, 2000 draws, horizons 4): accepted 165 -> 135, ESS 63.6 -> 65.3 GDP medians +0.05 to +0.11pp; CPI within +/-0.02pp Acceptance falls because the longer sample tightens the posterior, so the draw-count calibration in the comment and the forecast_uk docstring is re-measured on the new sample. 2000 remains the smallest default clearing the 100-accepted threshold (1000 gives 69); ESS >= 100 still needs ~3500. Also updates the five site locations quoting the old endpoint, so the published sample matches what the server reports. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
The guard hardcoded "estimated through 2023Q2", so it failed on exactly the deliberate refresh it exists to permit, while a literal also cannot catch drift to any other quarter. It now reads the constant and rejects any other quarter presented as the estimation endpoint. Verified non-vacuous: editing the site to a different quarter fails it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The hosted server reported two different samples for the same model:
model_summaryread the committed replication artifacts (1992Q1–2025Q1, refreshed in #85) whileforecast_ukandget_model_statusreported 1992Q1–2023Q2.That was not just a label —
_SVAR_EST_ENDtruncates the estimation frame in_estimate()(core.py:752), so the live forecast was discarding seven quarters of data the package already ships.The constant is deliberately pinned so hosted results only move on a deliberate refresh. This is that refresh. Held one quarter short of the 2026Q1 data edge, matching the repo, so the forecast origin stays outside the estimation sample.
Measured effect (seed 0, 2000 draws, horizons 4)
GDP medians shift +0.05 to +0.11pp; CPI stays within ±0.02pp. Small and coherent — no discontinuity.
Draw-count recalibration
Acceptance falls (~8% → ~6.8%) because the longer sample tightens the posterior, so the figures in the
_SVAR_DEFAULT_DRAWScomment and theforecast_ukdocstring were re-measured:2000 remains the smallest default clearing the 100-accepted threshold — 1000 would not (69). ESS ≥ 100 still needs ~3500.
Also updates the five site locations quoting the old endpoint, so the published sample matches what the server reports.
🤖 Generated with Claude Code