context7.json

{
  "$schema": "https://context7.com/schema/context7.json",
  "projectTitle": "tsbootstrap",
  "description": "Time series bootstrapping in Python: one typed bootstrap() entry point across IID, block, residual, sieve, and model-based (AR/ARIMA/VAR) methods, with an uncertainty-quantification layer (EnbPI, adaptive conformal ACI/NexCP, AR forecast intervals) and a diagnose() recommender.",
  "folders": ["docs/source"],
  "excludeFolders": ["docs/build", "tests", "htmlcov", "docs/source/_static"],
  "rules": [
    "The public entry point is bootstrap(X, *, method=...); typed method specs live in tsbootstrap.methods (IID, MovingBlock, CircularBlock, StationaryBlock, NonOverlappingBlock, TaperedBlock, ResidualBootstrap, SieveAR).",
    "Block length defaults to automatic Politis-White selection via block_length='auto' (avg_block_length='auto' for StationaryBlock).",
    "Model-based methods (AR/ARIMA/VAR residual, sieve) need the 'models' extra; the uncertainty-quantification layer tsbootstrap.uq needs the 'uq' extra.",
    "Call diagnose(X).recommended_methods for a programmatic method recommendation on a series.",
    "forecast_intervals currently supports AR models only."
  ]
}