Datasets:
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
tags:
|
| 6 |
+
- time-series
|
| 7 |
+
- forecasting
|
| 8 |
+
- chronos
|
| 9 |
+
- patch-aliasing
|
| 10 |
+
- bayesian
|
| 11 |
+
size_categories:
|
| 12 |
+
- 10K<n<100K
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# Patch-Aliasing Bayesian Analysis Data
|
| 16 |
+
|
| 17 |
+
Raw and processed data from the Bayesian analysis of structural patch-aliasing in Chronos-Bolt Tiny.
|
| 18 |
+
Companion to the model weights at [`federicosabbadini/chronos-bolt-patch-aliasing-models`](https://huggingface.co/federicosabbadini/chronos-bolt-patch-aliasing-models).
|
| 19 |
+
|
| 20 |
+
## Structure
|
| 21 |
+
|
| 22 |
+
```
|
| 23 |
+
clean_15model/ # 15 preregistered (P,S) configurations
|
| 24 |
+
full_22model/ # all 22 configurations (adds 7 robustness checks)
|
| 25 |
+
h1_fixed_offset/ # supplementary H1 analysis with fixed offset
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
Each run folder contains:
|
| 29 |
+
|
| 30 |
+
- `01_prior_spec.json` / `.parquet` — prior specification
|
| 31 |
+
- `03_recovery.parquet` — parameter recovery test
|
| 32 |
+
- `04_*.nc` — MCMC trace files (NetCDF/ArviZ InferenceData)
|
| 33 |
+
- `05_diagnostics.parquet` — convergence diagnostics
|
| 34 |
+
- `05_sensitivity.parquet` — prior sensitivity analysis
|
| 35 |
+
- `05_verdicts.parquet` — hypothesis verdicts
|
| 36 |
+
- `data/` — input data (contrasts, collapse profiles, MDL cells, sites)
|
| 37 |
+
- `data/signals/` — generated probe signals
|
| 38 |
+
- `figures/` — analysis figures (prior predictive, forests, combs, PPC, sensitivity)
|
| 39 |
+
- `run.log` — execution log
|
| 40 |
+
|
| 41 |
+
## Models fitted
|
| 42 |
+
|
| 43 |
+
| ID | Estimand | Hypothesis |
|
| 44 |
+
|---|---|---|
|
| 45 |
+
| A | Population forecast attenuation | H1 behavioural |
|
| 46 |
+
| B | Codelength expansion at locked frequencies | H1 representational |
|
| 47 |
+
| C | Phase-offset dispersion | H2 |
|
| 48 |
+
| D1 | Token-collapse grid labelling | H3a (stride branch) |
|
| 49 |
+
| D2 | Stride-scaling slope | H3a (κ_S) |
|
| 50 |
+
|
| 51 |
+
## Key results
|
| 52 |
+
|
| 53 |
+
- **H1 behavioural**: inconclusive (forecast contrast 1.08 [0.45, 2.49])
|
| 54 |
+
- **H1 representational**: supported on clean (codelength 1.27 [1.23, 1.31])
|
| 55 |
+
- **H2 phase invariance**: supported (σ_φ = 0.033, ROPE prob 0.997)
|
| 56 |
+
- **H3a stride branch**: supported (κ_S = 1.000 [0.992, 1.008])
|
| 57 |
+
- **H3b patch branch**: not identified
|
| 58 |
+
|
| 59 |
+
## Trace file sizes
|
| 60 |
+
|
| 61 |
+
The `.nc` files contain full MCMC traces (4 chains × 2000 draws). Models A and C are large (~8 GB each) due to per-observation random effects; Models B, D1, D2 are smaller (4–730 MB).
|
| 62 |
+
|
| 63 |
+
## License
|
| 64 |
+
|
| 65 |
+
Apache-2.0 — same as the Chronos models these data derive from.
|