Data & Provenance¶
Canonical data interface¶
import asrquant as asr
data = asr.data.load("prices.csv", date_column="Date")
quality = asr.data.validate(data)
asr.data.load accepts supported local table formats, public HTTP(S) tables and provider-backed data through the broader data layer. Existing helpers such as load_prices, load_sql, clean_prices, simple_returns and log_returns remain available.
Provider surface¶
The 1.2 package includes provider abstractions for ECB, Yahoo, FRED, Binance and Alpha Vantage. Availability, credentials and external revisions remain provider-specific.
Recommended research checks¶
Before downstream modelling, record:
- source and retrieval timestamp;
- timestamp semantics and timezone;
- publication/revision lag when relevant;
- missing-value policy;
- corporate-action or instrument adjustments;
- data fingerprint or equivalent reproducibility metadata.
Point-in-time principle¶
A final-vintage dataset can contain information that was not available to the strategy at historical decision time. For empirical research, reconstruct the information set that actually existed at each date whenever revisions or publication lags matter.